Merge remote-tracking branch 'upstream/master'

Conflicts:
	plugins/inputs/dovecot/README.md
	plugins/inputs/dovecot/dovecot.go
This commit is contained in:
Miki 2016-02-11 09:14:28 +01:00
commit 9e87128588
85 changed files with 3788 additions and 572 deletions

View File

@ -1,10 +1,21 @@
## v0.10.3 [unreleased]
### Release Notes
- Users of the `exec` and `kafka_consumer` can now specify the incoming data
format that they would like to parse. Currently supports: "json", "influx", and
"graphite"
- More info on parsing arbitrary data formats can be found
[here](https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md)
### Features
- [#652](https://github.com/influxdata/telegraf/pull/652): CouchDB Input Plugin
- [#655](https://github.com/influxdata/telegraf/pull/655): Support parsing arbitrary data formats. Currently limited to kafka_consumer and exec inputs.
- [#671](https://github.com/influxdata/telegraf/pull/671): Dovecot input plugin.
### Bugfixes
- [#443](https://github.com/influxdata/telegraf/issues/443): Fix Ping command timeout parameter on Linux.
- [#662](https://github.com/influxdata/telegraf/pull/667): Change `[tags]` to `[global_tags]` to fix multiple-plugin tags bug.
- [#642](https://github.com/influxdata/telegraf/issues/642): Riemann output plugin issues.
## v0.10.2 [2016-02-04]

View File

@ -9,9 +9,9 @@ To generate a file with specific inputs and outputs, you can use the
-input-filter and -output-filter flags:
`telegraf -sample-config -input-filter cpu:mem:net:swap -output-filter influxdb:kafka`
## `[tags]` Configuration
## `[global_tags]` Configuration
Global tags can be specific in the `[tags]` section of the config file in
Global tags can be specific in the `[global_tags]` section of the config file in
key="value" format. All metrics being gathered on this host will be tagged
with the tags specified here.
@ -76,7 +76,7 @@ measurements at a 10s interval and will collect per-cpu data, dropping any
fields which begin with `time_`.
```toml
[tags]
[global_tags]
dc = "denver-1"
[agent]

View File

@ -129,6 +129,52 @@ func init() {
}
```
## Input Plugins Accepting Arbitrary Data Formats
Some input plugins (such as
[exec](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/exec))
accept arbitrary input data formats. An overview of these data formats can
be found
[here](https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md).
In order to enable this, you must specify a `SetParser(parser parsers.Parser)`
function on the plugin object (see the exec plugin for an example), as well as
defining `parser` as a field of the object.
You can then utilize the parser internally in your plugin, parsing data as you
see fit. Telegraf's configuration layer will take care of instantiating and
creating the `Parser` object.
You should also add the following to your SampleConfig() return:
```toml
### Data format to consume. This can be "json", "influx" or "graphite"
### Each data format has it's own unique set of configuration options, read
### more about them here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md
data_format = "influx"
```
Below is the `Parser` interface.
```go
// Parser is an interface defining functions that a parser plugin must satisfy.
type Parser interface {
// Parse takes a byte buffer separated by newlines
// ie, `cpu.usage.idle 90\ncpu.usage.busy 10`
// and parses it into telegraf metrics
Parse(buf []byte) ([]telegraf.Metric, error)
// ParseLine takes a single string metric
// ie, "cpu.usage.idle 90"
// and parses it into a telegraf metric.
ParseLine(line string) (telegraf.Metric, error)
}
```
And you can view the code
[here.](https://github.com/influxdata/telegraf/blob/henrypfhu-master/plugins/parsers/registry.go)
## Service Input Plugins
This section is for developers who want to create new "service" collection

274
DATA_FORMATS_INPUT.md Normal file
View File

@ -0,0 +1,274 @@
# Telegraf Input Data Formats
Telegraf metrics, like InfluxDB
[points](https://docs.influxdata.com/influxdb/v0.10/write_protocols/line/),
are a combination of four basic parts:
1. Measurement Name
1. Tags
1. Fields
1. Timestamp
These four parts are easily defined when using InfluxDB line-protocol as a
data format. But there are other data formats that users may want to use which
require more advanced configuration to create usable Telegraf metrics.
Plugins such as `exec` and `kafka_consumer` parse textual data. Up until now,
these plugins were statically configured to parse just a single
data format. `exec` mostly only supported parsing JSON, and `kafka_consumer` only
supported data in InfluxDB line-protocol.
But now we are normalizing the parsing of various data formats across all
plugins that can support it. You will be able to identify a plugin that supports
different data formats by the presence of a `data_format` config option, for
example, in the exec plugin:
```toml
[[inputs.exec]]
### Commands array
commands = ["/tmp/test.sh", "/usr/bin/mycollector --foo=bar"]
### measurement name suffix (for separating different commands)
name_suffix = "_mycollector"
### Data format to consume. This can be "json", "influx" or "graphite"
### Each data format has it's own unique set of configuration options, read
### more about them here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md
data_format = "json"
### Additional configuration options go here
```
Each data_format has an additional set of configuration options available, which
I'll go over below.
## Influx:
There are no additional configuration options for InfluxDB line-protocol. The
metrics are parsed directly into Telegraf metrics.
#### Influx Configuration:
```toml
[[inputs.exec]]
### Commands array
commands = ["/tmp/test.sh", "/usr/bin/mycollector --foo=bar"]
### measurement name suffix (for separating different commands)
name_suffix = "_mycollector"
### Data format to consume. This can be "json", "influx" or "graphite"
### Each data format has it's own unique set of configuration options, read
### more about them here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md
data_format = "influx"
```
## JSON:
The JSON data format flattens JSON into metric _fields_. For example, this JSON:
```json
{
"a": 5,
"b": {
"c": 6
}
}
```
Would get translated into _fields_ of a measurement:
```
myjsonmetric a=5,b_c=6
```
The _measurement_ _name_ is usually the name of the plugin,
but can be overridden using the `name_override` config option.
#### JSON Configuration:
The JSON data format supports specifying "tag keys". If specified, keys
will be searched for in the root-level of the JSON blob. If the key(s) exist,
they will be applied as tags to the Telegraf metrics.
For example, if you had this configuration:
```toml
[[inputs.exec]]
### Commands array
commands = ["/tmp/test.sh", "/usr/bin/mycollector --foo=bar"]
### measurement name suffix (for separating different commands)
name_suffix = "_mycollector"
### Data format to consume. This can be "json", "influx" or "graphite"
### Each data format has it's own unique set of configuration options, read
### more about them here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md
data_format = "json"
### List of tag names to extract from top-level of JSON server response
tag_keys = [
"my_tag_1",
"my_tag_2"
]
```
with this JSON output from a command:
```json
{
"a": 5,
"b": {
"c": 6
},
"my_tag_1": "foo"
}
```
Your Telegraf metrics would get tagged with "my_tag_1"
```
exec_mycollector,my_tag_1=foo a=5,b_c=6
```
## Graphite:
The Graphite data format translates graphite _dot_ buckets directly into
telegraf measurement names, with a single value field, and without any tags. For
more advanced options, Telegraf supports specifying "templates" to translate
graphite buckets into Telegraf metrics.
#### Separator:
You can specify a separator to use for the parsed metrics.
By default, it will leave the metrics with a "." separator.
Setting `separator = "_"` will translate:
```
cpu.usage.idle 99
=> cpu_usage_idle value=99
```
#### Measurement/Tag Templates:
The most basic template is to specify a single transformation to apply to all
incoming metrics. _measurement_ is a special keyword that tells Telegraf which
parts of the graphite bucket to combine into the measurement name. It can have a
trailing `*` to indicate that the remainder of the metric should be used.
Other words are considered tag keys. So the following template:
```toml
templates = [
"region.measurement*"
]
```
would result in the following Graphite -> Telegraf transformation.
```
us-west.cpu.load 100
=> cpu.load,region=us-west value=100
```
#### Field Templates:
There is also a _field_ keyword, which can only be specified once.
The field keyword tells Telegraf to give the metric that field name.
So the following template:
```toml
templates = [
"measurement.measurement.field.region"
]
```
would result in the following Graphite -> Telegraf transformation.
```
cpu.usage.idle.us-west 100
=> cpu_usage,region=us-west idle=100
```
#### Filter Templates:
Users can also filter the template(s) to use based on the name of the bucket,
using glob matching, like so:
```toml
templates = [
"cpu.* measurement.measurement.region",
"mem.* measurement.measurement.host"
]
```
which would result in the following transformation:
```
cpu.load.us-west 100
=> cpu_load,region=us-west value=100
mem.cached.localhost 256
=> mem_cached,host=localhost value=256
```
#### Adding Tags:
Additional tags can be added to a metric that don't exist on the received metric.
You can add additional tags by specifying them after the pattern.
Tags have the same format as the line protocol.
Multiple tags are separated by commas.
```toml
templates = [
"measurement.measurement.field.region datacenter=1a"
]
```
would result in the following Graphite -> Telegraf transformation.
```
cpu.usage.idle.us-west 100
=> cpu_usage,region=us-west,datacenter=1a idle=100
```
There are many more options available,
[More details can be found here](https://github.com/influxdata/influxdb/tree/master/services/graphite#templates)
#### Graphite Configuration:
```toml
[[inputs.exec]]
### Commands array
commands = ["/tmp/test.sh", "/usr/bin/mycollector --foo=bar"]
### measurement name suffix (for separating different commands)
name_suffix = "_mycollector"
### Data format to consume. This can be "json", "influx" or "graphite" (line-protocol)
### Each data format has it's own unique set of configuration options, read
### more about them here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md
data_format = "graphite"
### This string will be used to join the matched values.
separator = "_"
### Each template line requires a template pattern. It can have an optional
### filter before the template and separated by spaces. It can also have optional extra
### tags following the template. Multiple tags should be separated by commas and no spaces
### similar to the line protocol format. There can be only one default template.
### Templates support below format:
### 1. filter + template
### 2. filter + template + extra tag
### 3. filter + template with field key
### 4. default template
templates = [
"*.app env.service.resource.measurement",
"stats.* .host.measurement* region=us-west,agent=sensu",
"stats2.* .host.measurement.field",
"measurement*"
]
```

12
Godeps
View File

@ -2,10 +2,8 @@ git.eclipse.org/gitroot/paho/org.eclipse.paho.mqtt.golang.git dbd8d5c40a582eb9ad
github.com/Shopify/sarama d37c73f2b2bce85f7fa16b6a550d26c5372892ef
github.com/Sirupsen/logrus f7f79f729e0fbe2fcc061db48a9ba0263f588252
github.com/amir/raidman 6a8e089bbe32e6b907feae5ba688841974b3c339
github.com/armon/go-metrics 345426c77237ece5dab0e1605c3e4b35c3f54757
github.com/aws/aws-sdk-go 87b1e60a50b09e4812dee560b33a238f67305804
github.com/beorn7/perks b965b613227fddccbfffe13eae360ed3fa822f8d
github.com/boltdb/bolt ee4a0888a9abe7eefe5a0992ca4cb06864839873
github.com/cenkalti/backoff 4dc77674aceaabba2c7e3da25d4c823edfb73f99
github.com/dancannon/gorethink 6f088135ff288deb9d5546f4c71919207f891a70
github.com/davecgh/go-spew 5215b55f46b2b919f50a1df0eaa5886afe4e3b3d
@ -14,19 +12,15 @@ github.com/eapache/queue ded5959c0d4e360646dc9e9908cff48666781367
github.com/fsouza/go-dockerclient 7b651349f9479f5114913eefbfd3c4eeddd79ab4
github.com/go-ini/ini afbd495e5aaea13597b5e14fe514ddeaa4d76fc3
github.com/go-sql-driver/mysql 7c7f556282622f94213bc028b4d0a7b6151ba239
github.com/gogo/protobuf e8904f58e872a473a5b91bc9bf3377d223555263
github.com/golang/protobuf 6aaa8d47701fa6cf07e914ec01fde3d4a1fe79c3
github.com/golang/snappy 723cc1e459b8eea2dea4583200fd60757d40097a
github.com/gonuts/go-shellquote e842a11b24c6abfb3dd27af69a17f482e4b483c2
github.com/gorilla/context 1c83b3eabd45b6d76072b66b746c20815fb2872d
github.com/gorilla/mux 26a6070f849969ba72b72256e9f14cf519751690
github.com/hailocab/go-hostpool e80d13ce29ede4452c43dea11e79b9bc8a15b478
github.com/hashicorp/go-msgpack fa3f63826f7c23912c15263591e65d54d080b458
github.com/hashicorp/raft 057b893fd996696719e98b6c44649ea14968c811
github.com/hashicorp/raft-boltdb d1e82c1ec3f15ee991f7cc7ffd5b67ff6f5bbaee
github.com/influxdata/config bae7cb98197d842374d3b8403905924094930f24
github.com/influxdata/influxdb 697f48b4e62e514e701ffec39978b864a3c666e6
github.com/influxdb/influxdb 697f48b4e62e514e701ffec39978b864a3c666e6
github.com/influxdata/influxdb a9552fdd91361819a792f337e5d9998859732a67
github.com/influxdb/influxdb a9552fdd91361819a792f337e5d9998859732a67
github.com/jmespath/go-jmespath c01cf91b011868172fdcd9f41838e80c9d716264
github.com/klauspost/crc32 999f3125931f6557b991b2f8472172bdfa578d38
github.com/lib/pq 8ad2b298cadd691a77015666a5372eae5dbfac8f
@ -42,7 +36,7 @@ github.com/prometheus/client_model fa8ad6fec33561be4280a8f0514318c79d7f6cb6
github.com/prometheus/common 14ca1097bbe21584194c15e391a9dab95ad42a59
github.com/prometheus/procfs 406e5b7bfd8201a36e2bb5f7bdae0b03380c2ce8
github.com/samuel/go-zookeeper 218e9c81c0dd8b3b18172b2bbfad92cc7d6db55f
github.com/shirou/gopsutil 85bf0974ed06e4e668595ae2b4de02e772a2819b
github.com/shirou/gopsutil e77438504d45b9985c99a75730fe65220ceea00e
github.com/soniah/gosnmp b1b4f885b12c5dcbd021c5cee1c904110de6db7d
github.com/streadway/amqp b4f3ceab0337f013208d31348b578d83c0064744
github.com/stretchr/objx 1a9d0bb9f541897e62256577b352fdbc1fb4fd94

View File

@ -1,34 +1,28 @@
git.eclipse.org/gitroot/paho/org.eclipse.paho.mqtt.golang.git dbd8d5c40a582eb9adacde36b47932b3a3ad0034
github.com/Shopify/sarama b1da1753dedcf77d053613b7eae907b98a2ddad5
github.com/Shopify/sarama d37c73f2b2bce85f7fa16b6a550d26c5372892ef
github.com/Sirupsen/logrus f7f79f729e0fbe2fcc061db48a9ba0263f588252
github.com/StackExchange/wmi f3e2bae1e0cb5aef83e319133eabfee30013a4a5
github.com/amir/raidman 6a8e089bbe32e6b907feae5ba688841974b3c339
github.com/armon/go-metrics 345426c77237ece5dab0e1605c3e4b35c3f54757
github.com/aws/aws-sdk-go 2a34ea8812f32aae75b43400f9424a0559840659
github.com/aws/aws-sdk-go 87b1e60a50b09e4812dee560b33a238f67305804
github.com/beorn7/perks b965b613227fddccbfffe13eae360ed3fa822f8d
github.com/boltdb/bolt ee4a0888a9abe7eefe5a0992ca4cb06864839873
github.com/cenkalti/backoff 4dc77674aceaabba2c7e3da25d4c823edfb73f99
github.com/dancannon/gorethink 6f088135ff288deb9d5546f4c71919207f891a70
github.com/davecgh/go-spew 5215b55f46b2b919f50a1df0eaa5886afe4e3b3d
github.com/eapache/go-resiliency b86b1ec0dd4209a588dc1285cdd471e73525c0b3
github.com/eapache/queue ded5959c0d4e360646dc9e9908cff48666781367
github.com/fsouza/go-dockerclient 02a8beb401b20e112cff3ea740545960b667eab1
github.com/fsouza/go-dockerclient 7b651349f9479f5114913eefbfd3c4eeddd79ab4
github.com/go-ini/ini afbd495e5aaea13597b5e14fe514ddeaa4d76fc3
github.com/go-ole/go-ole 50055884d646dd9434f16bbb5c9801749b9bafe4
github.com/go-sql-driver/mysql 7c7f556282622f94213bc028b4d0a7b6151ba239
github.com/gogo/protobuf e8904f58e872a473a5b91bc9bf3377d223555263
github.com/golang/protobuf 45bba206dd5270d96bac4942dcfe515726613249
github.com/golang/snappy 1963d058044b19e16595f80d5050fa54e2070438
github.com/golang/protobuf 6aaa8d47701fa6cf07e914ec01fde3d4a1fe79c3
github.com/golang/snappy 723cc1e459b8eea2dea4583200fd60757d40097a
github.com/gonuts/go-shellquote e842a11b24c6abfb3dd27af69a17f482e4b483c2
github.com/gorilla/context 1c83b3eabd45b6d76072b66b746c20815fb2872d
github.com/gorilla/mux 26a6070f849969ba72b72256e9f14cf519751690
github.com/hailocab/go-hostpool e80d13ce29ede4452c43dea11e79b9bc8a15b478
github.com/hashicorp/go-msgpack fa3f63826f7c23912c15263591e65d54d080b458
github.com/hashicorp/raft 057b893fd996696719e98b6c44649ea14968c811
github.com/hashicorp/raft-boltdb d1e82c1ec3f15ee991f7cc7ffd5b67ff6f5bbaee
github.com/influxdata/config bae7cb98197d842374d3b8403905924094930f24
github.com/influxdata/influxdb 60df13fb566d07ff2cdd07aa23a4796a02b0df3c
github.com/influxdb/influxdb 60df13fb566d07ff2cdd07aa23a4796a02b0df3c
github.com/influxdata/influxdb a9552fdd91361819a792f337e5d9998859732a67
github.com/influxdb/influxdb a9552fdd91361819a792f337e5d9998859732a67
github.com/jmespath/go-jmespath c01cf91b011868172fdcd9f41838e80c9d716264
github.com/klauspost/crc32 999f3125931f6557b991b2f8472172bdfa578d38
github.com/lib/pq 8ad2b298cadd691a77015666a5372eae5dbfac8f
@ -45,7 +39,7 @@ github.com/prometheus/client_model fa8ad6fec33561be4280a8f0514318c79d7f6cb6
github.com/prometheus/common 14ca1097bbe21584194c15e391a9dab95ad42a59
github.com/prometheus/procfs 406e5b7bfd8201a36e2bb5f7bdae0b03380c2ce8
github.com/samuel/go-zookeeper 218e9c81c0dd8b3b18172b2bbfad92cc7d6db55f
github.com/shirou/gopsutil 9d8191d6a6e17dcf43b10a20084a11e8c1aa92e6
github.com/shirou/gopsutil e77438504d45b9985c99a75730fe65220ceea00e
github.com/shirou/w32 ada3ba68f000aa1b58580e45c9d308fe0b7fc5c5
github.com/soniah/gosnmp b1b4f885b12c5dcbd021c5cee1c904110de6db7d
github.com/streadway/amqp b4f3ceab0337f013208d31348b578d83c0064744
@ -54,9 +48,8 @@ github.com/stretchr/testify f390dcf405f7b83c997eac1b06768bb9f44dec18
github.com/wvanbergen/kafka 1a8639a45164fcc245d5c7b4bd3ccfbd1a0ffbf3
github.com/wvanbergen/kazoo-go 0f768712ae6f76454f987c3356177e138df258f8
github.com/zensqlmonitor/go-mssqldb ffe5510c6fa5e15e6d983210ab501c815b56b363
golang.org/x/crypto 1f22c0103821b9390939b6776727195525381532
golang.org/x/net 04b9de9b512f58addf28c9853d50ebef61c3953e
golang.org/x/text 6fc2e00a0d64b1f7fc1212dae5b0c939cf6d9ac4
golang.org/x/text 6d3c22c4525a4da167968fa2479be5524d2e8bd0
gopkg.in/dancannon/gorethink.v1 6f088135ff288deb9d5546f4c71919207f891a70
gopkg.in/fatih/pool.v2 cba550ebf9bce999a02e963296d4bc7a486cb715
gopkg.in/mgo.v2 03c9f3ee4c14c8e51ee521a6a7d0425658dd6f64

View File

@ -155,10 +155,12 @@ Currently implemented sources:
* aerospike
* apache
* bcache
* couchdb
* disque
* docker
* dovecot
* elasticsearch
* exec (generic JSON-emitting executable plugin)
* exec (generic executable plugin, support JSON, influx and graphite)
* haproxy
* httpjson (generic JSON-emitting http service plugin)
* influxdb

View File

@ -412,7 +412,7 @@ def generate_md5_from_file(path):
m.update(data)
return m.hexdigest()
def build_packages(build_output, version, nightly=False, rc=None, iteration=1):
def build_packages(build_output, version, pkg_arch, nightly=False, rc=None, iteration=1):
outfiles = []
tmp_build_dir = create_temp_dir()
if debug:
@ -461,6 +461,9 @@ def build_packages(build_output, version, nightly=False, rc=None, iteration=1):
current_location = os.path.join(current_location, name + '.tar.gz')
if rc is not None:
package_iteration = "0.rc{}".format(rc)
saved_a = a
if pkg_arch is not None:
a = pkg_arch
if a == '386':
a = 'i386'
fpm_command = "fpm {} --name {} -a {} -t {} --version {} --iteration {} -C {} -p {} ".format(
@ -472,6 +475,8 @@ def build_packages(build_output, version, nightly=False, rc=None, iteration=1):
package_iteration,
build_root,
current_location)
if pkg_arch is not None:
a = saved_a
if package_type == "rpm":
fpm_command += "--depends coreutils "
fpm_command += "--depends lsof"
@ -506,6 +511,7 @@ def print_usage():
print("\t --goarm=<arm version> \n\t\t- Build for specified ARM version (when building for ARM). Default value is: 6")
print("\t --platform=<platform> \n\t\t- Build for specified platform. Acceptable values: linux, windows, darwin, or all")
print("\t --version=<version> \n\t\t- Version information to apply to build metadata. If not specified, will be pulled from repo tag.")
print("\t --pkgarch=<package-arch> \n\t\t- Package architecture if different from <arch>")
print("\t --commit=<commit> \n\t\t- Use specific commit for build (currently a NOOP).")
print("\t --branch=<branch> \n\t\t- Build from a specific branch (currently a NOOP).")
print("\t --rc=<rc number> \n\t\t- Whether or not the build is a release candidate (affects version information).")
@ -532,6 +538,7 @@ def main():
commit = None
target_platform = None
target_arch = None
package_arch = None
nightly = False
race = False
branch = None
@ -570,6 +577,9 @@ def main():
elif '--version' in arg:
# Version to assign to this build (0.9.5, etc)
version = arg.split("=")[1]
elif '--pkgarch' in arg:
# Package architecture if different from <arch> (armhf, etc)
package_arch = arg.split("=")[1]
elif '--rc' in arg:
# Signifies that this is a release candidate build.
rc = arg.split("=")[1]
@ -703,7 +713,7 @@ def main():
if not check_path_for("fpm"):
print("!! Cannot package without command 'fpm'. Stopping.")
return 1
packages = build_packages(build_output, version, nightly=nightly, rc=rc, iteration=iteration)
packages = build_packages(build_output, version, package_arch, nightly=nightly, rc=rc, iteration=iteration)
# Optionally upload to S3
if upload:
upload_packages(packages, bucket_name=upload_bucket, nightly=nightly)

View File

@ -11,6 +11,7 @@ import (
"github.com/influxdata/telegraf/agent"
"github.com/influxdata/telegraf/internal/config"
_ "github.com/influxdata/telegraf/plugins/inputs/all"
_ "github.com/influxdata/telegraf/plugins/outputs/all"
)

View File

@ -10,7 +10,7 @@
# file would generate.
# Global tags can be specified here in key="value" format.
[tags]
[global_tags]
# dc = "us-east-1" # will tag all metrics with dc=us-east-1
# rack = "1a"

View File

@ -15,6 +15,7 @@ import (
"github.com/influxdata/telegraf/internal/models"
"github.com/influxdata/telegraf/plugins/inputs"
"github.com/influxdata/telegraf/plugins/outputs"
"github.com/influxdata/telegraf/plugins/parsers"
"github.com/influxdata/config"
"github.com/naoina/toml/ast"
@ -126,7 +127,9 @@ func (c *Config) ListTags() string {
return strings.Join(tags, " ")
}
var header = `# Telegraf configuration
var header = `###############################################################################
# Telegraf Configuration #
###############################################################################
# Telegraf is entirely plugin driven. All metrics are gathered from the
# declared inputs, and sent to the declared outputs.
@ -138,41 +141,41 @@ var header = `# Telegraf configuration
# file would generate.
# Global tags can be specified here in key="value" format.
[tags]
[global_tags]
# dc = "us-east-1" # will tag all metrics with dc=us-east-1
# rack = "1a"
# Configuration for telegraf agent
[agent]
# Default data collection interval for all inputs
### Default data collection interval for all inputs
interval = "10s"
# Rounds collection interval to 'interval'
# ie, if interval="10s" then always collect on :00, :10, :20, etc.
### Rounds collection interval to 'interval'
### ie, if interval="10s" then always collect on :00, :10, :20, etc.
round_interval = true
# Telegraf will cache metric_buffer_limit metrics for each output, and will
# flush this buffer on a successful write.
### Telegraf will cache metric_buffer_limit metrics for each output, and will
### flush this buffer on a successful write.
metric_buffer_limit = 10000
# Collection jitter is used to jitter the collection by a random amount.
# Each plugin will sleep for a random time within jitter before collecting.
# This can be used to avoid many plugins querying things like sysfs at the
# same time, which can have a measurable effect on the system.
### Collection jitter is used to jitter the collection by a random amount.
### Each plugin will sleep for a random time within jitter before collecting.
### This can be used to avoid many plugins querying things like sysfs at the
### same time, which can have a measurable effect on the system.
collection_jitter = "0s"
# Default data flushing interval for all outputs. You should not set this below
# interval. Maximum flush_interval will be flush_interval + flush_jitter
### Default flushing interval for all outputs. You shouldn't set this below
### interval. Maximum flush_interval will be flush_interval + flush_jitter
flush_interval = "10s"
# Jitter the flush interval by a random amount. This is primarily to avoid
# large write spikes for users running a large number of telegraf instances.
# ie, a jitter of 5s and interval 10s means flushes will happen every 10-15s
### Jitter the flush interval by a random amount. This is primarily to avoid
### large write spikes for users running a large number of telegraf instances.
### ie, a jitter of 5s and interval 10s means flushes will happen every 10-15s
flush_jitter = "0s"
# Run telegraf in debug mode
### Run telegraf in debug mode
debug = false
# Run telegraf in quiet mode
### Run telegraf in quiet mode
quiet = false
# Override default hostname, if empty use os.Hostname()
### Override default hostname, if empty use os.Hostname()
hostname = ""
@ -333,9 +336,9 @@ func (c *Config) LoadConfig(path string) error {
log.Printf("Could not parse [agent] config\n")
return err
}
case "tags":
case "global_tags", "tags":
if err = config.UnmarshalTable(subTable, c.Tags); err != nil {
log.Printf("Could not parse [tags] config\n")
log.Printf("Could not parse [global_tags] config\n")
return err
}
case "outputs":
@ -428,6 +431,17 @@ func (c *Config) addInput(name string, table *ast.Table) error {
}
input := creator()
// If the input has a SetParser function, then this means it can accept
// arbitrary types of input, so build the parser and set it.
switch t := input.(type) {
case parsers.ParserInput:
parser, err := buildParser(name, table)
if err != nil {
return err
}
t.SetParser(parser)
}
pluginConfig, err := buildInput(name, table)
if err != nil {
return err
@ -583,6 +597,69 @@ func buildInput(name string, tbl *ast.Table) (*internal_models.InputConfig, erro
return cp, nil
}
// buildParser grabs the necessary entries from the ast.Table for creating
// a parsers.Parser object, and creates it, which can then be added onto
// an Input object.
func buildParser(name string, tbl *ast.Table) (parsers.Parser, error) {
c := &parsers.Config{}
if node, ok := tbl.Fields["data_format"]; ok {
if kv, ok := node.(*ast.KeyValue); ok {
if str, ok := kv.Value.(*ast.String); ok {
c.DataFormat = str.Value
}
}
}
// Legacy support, exec plugin originally parsed JSON by default.
if name == "exec" && c.DataFormat == "" {
c.DataFormat = "json"
} else if c.DataFormat == "" {
c.DataFormat = "influx"
}
if node, ok := tbl.Fields["separator"]; ok {
if kv, ok := node.(*ast.KeyValue); ok {
if str, ok := kv.Value.(*ast.String); ok {
c.Separator = str.Value
}
}
}
if node, ok := tbl.Fields["templates"]; ok {
if kv, ok := node.(*ast.KeyValue); ok {
if ary, ok := kv.Value.(*ast.Array); ok {
for _, elem := range ary.Value {
if str, ok := elem.(*ast.String); ok {
c.Templates = append(c.Templates, str.Value)
}
}
}
}
}
if node, ok := tbl.Fields["tag_keys"]; ok {
if kv, ok := node.(*ast.KeyValue); ok {
if ary, ok := kv.Value.(*ast.Array); ok {
for _, elem := range ary.Value {
if str, ok := elem.(*ast.String); ok {
c.TagKeys = append(c.TagKeys, str.Value)
}
}
}
}
}
c.MetricName = name
delete(tbl.Fields, "data_format")
delete(tbl.Fields, "separator")
delete(tbl.Fields, "templates")
delete(tbl.Fields, "tag_keys")
return parsers.NewParser(c)
}
// buildOutput parses output specific items from the ast.Table, builds the filter and returns an
// internal_models.OutputConfig to be inserted into internal_models.RunningInput
// Note: error exists in the return for future calls that might require error

View File

@ -9,6 +9,7 @@ import (
"github.com/influxdata/telegraf/plugins/inputs/exec"
"github.com/influxdata/telegraf/plugins/inputs/memcached"
"github.com/influxdata/telegraf/plugins/inputs/procstat"
"github.com/influxdata/telegraf/plugins/parsers"
"github.com/stretchr/testify/assert"
)
@ -91,6 +92,9 @@ func TestConfig_LoadDirectory(t *testing.T) {
"Testdata did not produce correct memcached metadata.")
ex := inputs.Inputs["exec"]().(*exec.Exec)
p, err := parsers.NewJSONParser("exec", nil, nil)
assert.NoError(t, err)
ex.SetParser(p)
ex.Command = "/usr/bin/myothercollector --foo=bar"
eConfig := &internal_models.InputConfig{
Name: "exec",

View File

@ -20,7 +20,7 @@
# with 'required'. Be sure to edit those to make this configuration work.
# Tags can also be specified via a normal map, but only one form at a time:
[tags]
[global_tags]
dc = "us-east-1"
# Configuration for telegraf agent

View File

@ -9,7 +9,6 @@ import (
"fmt"
"io/ioutil"
"os"
"strconv"
"strings"
"time"
)
@ -35,47 +34,6 @@ func (d *Duration) UnmarshalTOML(b []byte) error {
var NotImplementedError = errors.New("not implemented yet")
type JSONFlattener struct {
Fields map[string]interface{}
}
// FlattenJSON flattens nested maps/interfaces into a fields map
func (f *JSONFlattener) FlattenJSON(
fieldname string,
v interface{},
) error {
if f.Fields == nil {
f.Fields = make(map[string]interface{})
}
fieldname = strings.Trim(fieldname, "_")
switch t := v.(type) {
case map[string]interface{}:
for k, v := range t {
err := f.FlattenJSON(fieldname+"_"+k+"_", v)
if err != nil {
return err
}
}
case []interface{}:
for i, v := range t {
k := strconv.Itoa(i)
err := f.FlattenJSON(fieldname+"_"+k+"_", v)
if err != nil {
return nil
}
}
case float64:
f.Fields[fieldname] = t
case bool, string, nil:
// ignored types
return nil
default:
return fmt.Errorf("JSON Flattener: got unexpected type %T with value %v (%s)",
t, t, fieldname)
}
return nil
}
// ReadLines reads contents from a file and splits them by new lines.
// A convenience wrapper to ReadLinesOffsetN(filename, 0, -1).
func ReadLines(filename string) ([]string, error) {

View File

@ -1,11 +1,9 @@
package telegraf
import (
"bytes"
"time"
"github.com/influxdata/influxdb/client/v2"
"github.com/influxdata/influxdb/models"
)
type Metric interface {
@ -63,25 +61,6 @@ func NewMetric(
}, nil
}
// ParseMetrics returns a slice of Metrics from a text representation of a
// metric (in line-protocol format)
// with each metric separated by newlines. If any metrics fail to parse,
// a non-nil error will be returned in addition to the metrics that parsed
// successfully.
func ParseMetrics(buf []byte) ([]Metric, error) {
// parse even if the buffer begins with a newline
buf = bytes.TrimPrefix(buf, []byte("\n"))
points, err := models.ParsePoints(buf)
metrics := make([]Metric, len(points))
for i, point := range points {
// Ignore error here because it's impossible that a model.Point
// wouldn't parse into client.Point properly
metrics[i], _ = NewMetric(point.Name(), point.Tags(),
point.Fields(), point.Time())
}
return metrics, err
}
func (m *metric) Name() string {
return m.pt.Name()
}

View File

@ -9,58 +9,6 @@ import (
"github.com/stretchr/testify/assert"
)
const validMs = `
cpu,cpu=cpu0,host=foo,datacenter=us-east usage_idle=99,usage_busy=1 1454105876344540456
`
const invalidMs = `
cpu, cpu=cpu0,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo usage_idle
cpu,host usage_idle=99
cpu,host=foo usage_idle=99 very bad metric
`
const validInvalidMs = `
cpu,cpu=cpu0,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,cpu=cpu1,host=foo,datacenter=us-east usage_idle=51,usage_busy=49
cpu,cpu=cpu2,host=foo,datacenter=us-east usage_idle=60,usage_busy=40
cpu,host usage_idle=99
`
func TestParseValidMetrics(t *testing.T) {
metrics, err := ParseMetrics([]byte(validMs))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
m := metrics[0]
tags := map[string]string{
"host": "foo",
"datacenter": "us-east",
"cpu": "cpu0",
}
fields := map[string]interface{}{
"usage_idle": float64(99),
"usage_busy": float64(1),
}
assert.Equal(t, tags, m.Tags())
assert.Equal(t, fields, m.Fields())
assert.Equal(t, "cpu", m.Name())
assert.Equal(t, int64(1454105876344540456), m.UnixNano())
}
func TestParseInvalidMetrics(t *testing.T) {
metrics, err := ParseMetrics([]byte(invalidMs))
assert.Error(t, err)
assert.Len(t, metrics, 0)
}
func TestParseValidAndInvalidMetrics(t *testing.T) {
metrics, err := ParseMetrics([]byte(validInvalidMs))
assert.Error(t, err)
assert.Len(t, metrics, 3)
}
func TestNewMetric(t *testing.T) {
now := time.Now()

View File

@ -104,11 +104,9 @@ type Aerospike struct {
}
var sampleConfig = `
# Aerospike servers to connect to (with port)
# Default: servers = ["localhost:3000"]
#
# This plugin will query all namespaces the aerospike
# server has configured and get stats for them.
### Aerospike servers to connect to (with port)
### This plugin will query all namespaces the aerospike
### server has configured and get stats for them.
servers = ["localhost:3000"]
`

View File

@ -4,6 +4,7 @@ import (
_ "github.com/influxdata/telegraf/plugins/inputs/aerospike"
_ "github.com/influxdata/telegraf/plugins/inputs/apache"
_ "github.com/influxdata/telegraf/plugins/inputs/bcache"
_ "github.com/influxdata/telegraf/plugins/inputs/couchdb"
_ "github.com/influxdata/telegraf/plugins/inputs/disque"
_ "github.com/influxdata/telegraf/plugins/inputs/docker"
_ "github.com/influxdata/telegraf/plugins/inputs/dovecot"

View File

@ -20,7 +20,7 @@ type Apache struct {
}
var sampleConfig = `
# An array of Apache status URI to gather stats.
### An array of Apache status URI to gather stats.
urls = ["http://localhost/server-status?auto"]
`

View File

@ -18,14 +18,14 @@ type Bcache struct {
}
var sampleConfig = `
# Bcache sets path
# If not specified, then default is:
# bcachePath = "/sys/fs/bcache"
#
# By default, telegraf gather stats for all bcache devices
# Setting devices will restrict the stats to the specified
# bcache devices.
# bcacheDevs = ["bcache0", ...]
### Bcache sets path
### If not specified, then default is:
bcachePath = "/sys/fs/bcache"
### By default, telegraf gather stats for all bcache devices
### Setting devices will restrict the stats to the specified
### bcache devices.
bcacheDevs = ["bcache0"]
`
func (b *Bcache) SampleConfig() string {

View File

@ -0,0 +1,255 @@
# CouchDB Input Plugin
---
The CouchDB plugin gathers metrics of CouchDB using [_stats](http://docs.couchdb.org/en/1.6.1/api/server/common.html?highlight=stats#get--_stats) endpoint.
### Configuration:
```
# Sample Config:
[[inputs.couchdb]]
hosts = ["http://localhost:5984/_stats"]
```
### Measurements & Fields:
Statistics specific to the internals of CouchDB:
- couchdb_auth_cache_misses
- couchdb_database_writes
- couchdb_open_databases
- couchdb_auth_cache_hits
- couchdb_request_time
- couchdb_database_reads
- couchdb_open_os_files
Statistics of HTTP requests by method:
- httpd_request_methods_put
- httpd_request_methods_get
- httpd_request_methods_copy
- httpd_request_methods_delete
- httpd_request_methods_post
- httpd_request_methods_head
Statistics of HTTP requests by response code:
- httpd_status_codes_200
- httpd_status_codes_201
- httpd_status_codes_202
- httpd_status_codes_301
- httpd_status_codes_304
- httpd_status_codes_400
- httpd_status_codes_401
- httpd_status_codes_403
- httpd_status_codes_404
- httpd_status_codes_405
- httpd_status_codes_409
- httpd_status_codes_412
- httpd_status_codes_500
httpd statistics:
- httpd_clients_requesting_changes
- httpd_temporary_view_reads
- httpd_requests
- httpd_bulk_requests
- httpd_view_reads
### Tags:
- server (url of the couchdb _stats endpoint)
### Example output:
```
➜ telegraf git:(master) ✗ ./telegraf -config ./config.conf -input-filter couchdb -test
* Plugin: couchdb,
Collection 1
> couchdb,server=http://localhost:5984/_stats couchdb_auth_cache_hits_current=0,
couchdb_auth_cache_hits_max=0,
couchdb_auth_cache_hits_mean=0,
couchdb_auth_cache_hits_min=0,
couchdb_auth_cache_hits_stddev=0,
couchdb_auth_cache_hits_sum=0,
couchdb_auth_cache_misses_current=0,
couchdb_auth_cache_misses_max=0,
couchdb_auth_cache_misses_mean=0,
couchdb_auth_cache_misses_min=0,
couchdb_auth_cache_misses_stddev=0,
couchdb_auth_cache_misses_sum=0,
couchdb_database_reads_current=0,
couchdb_database_reads_max=0,
couchdb_database_reads_mean=0,
couchdb_database_reads_min=0,
couchdb_database_reads_stddev=0,
couchdb_database_reads_sum=0,
couchdb_database_writes_current=1102,
couchdb_database_writes_max=131,
couchdb_database_writes_mean=0.116,
couchdb_database_writes_min=0,
couchdb_database_writes_stddev=3.536,
couchdb_database_writes_sum=1102,
couchdb_open_databases_current=1,
couchdb_open_databases_max=1,
couchdb_open_databases_mean=0,
couchdb_open_databases_min=0,
couchdb_open_databases_stddev=0.01,
couchdb_open_databases_sum=1,
couchdb_open_os_files_current=2,
couchdb_open_os_files_max=2,
couchdb_open_os_files_mean=0,
couchdb_open_os_files_min=0,
couchdb_open_os_files_stddev=0.02,
couchdb_open_os_files_sum=2,
couchdb_request_time_current=242.21,
couchdb_request_time_max=102,
couchdb_request_time_mean=5.767,
couchdb_request_time_min=1,
couchdb_request_time_stddev=17.369,
couchdb_request_time_sum=242.21,
httpd_bulk_requests_current=0,
httpd_bulk_requests_max=0,
httpd_bulk_requests_mean=0,
httpd_bulk_requests_min=0,
httpd_bulk_requests_stddev=0,
httpd_bulk_requests_sum=0,
httpd_clients_requesting_changes_current=0,
httpd_clients_requesting_changes_max=0,
httpd_clients_requesting_changes_mean=0,
httpd_clients_requesting_changes_min=0,
httpd_clients_requesting_changes_stddev=0,
httpd_clients_requesting_changes_sum=0,
httpd_request_methods_copy_current=0,
httpd_request_methods_copy_max=0,
httpd_request_methods_copy_mean=0,
httpd_request_methods_copy_min=0,
httpd_request_methods_copy_stddev=0,
httpd_request_methods_copy_sum=0,
httpd_request_methods_delete_current=0,
httpd_request_methods_delete_max=0,
httpd_request_methods_delete_mean=0,
httpd_request_methods_delete_min=0,
httpd_request_methods_delete_stddev=0,
httpd_request_methods_delete_sum=0,
httpd_request_methods_get_current=31,
httpd_request_methods_get_max=1,
httpd_request_methods_get_mean=0.003,
httpd_request_methods_get_min=0,
httpd_request_methods_get_stddev=0.057,
httpd_request_methods_get_sum=31,
httpd_request_methods_head_current=0,
httpd_request_methods_head_max=0,
httpd_request_methods_head_mean=0,
httpd_request_methods_head_min=0,
httpd_request_methods_head_stddev=0,
httpd_request_methods_head_sum=0,
httpd_request_methods_post_current=1102,
httpd_request_methods_post_max=131,
httpd_request_methods_post_mean=0.116,
httpd_request_methods_post_min=0,
httpd_request_methods_post_stddev=3.536,
httpd_request_methods_post_sum=1102,
httpd_request_methods_put_current=1,
httpd_request_methods_put_max=1,
httpd_request_methods_put_mean=0,
httpd_request_methods_put_min=0,
httpd_request_methods_put_stddev=0.01,
httpd_request_methods_put_sum=1,
httpd_requests_current=1133,
httpd_requests_max=130,
httpd_requests_mean=0.118,
httpd_requests_min=0,
httpd_requests_stddev=3.512,
httpd_requests_sum=1133,
httpd_status_codes_200_current=31,
httpd_status_codes_200_max=1,
httpd_status_codes_200_mean=0.003,
httpd_status_codes_200_min=0,
httpd_status_codes_200_stddev=0.057,
httpd_status_codes_200_sum=31,
httpd_status_codes_201_current=1103,
httpd_status_codes_201_max=130,
httpd_status_codes_201_mean=0.116,
httpd_status_codes_201_min=0,
httpd_status_codes_201_stddev=3.532,
httpd_status_codes_201_sum=1103,
httpd_status_codes_202_current=0,
httpd_status_codes_202_max=0,
httpd_status_codes_202_mean=0,
httpd_status_codes_202_min=0,
httpd_status_codes_202_stddev=0,
httpd_status_codes_202_sum=0,
httpd_status_codes_301_current=0,
httpd_status_codes_301_max=0,
httpd_status_codes_301_mean=0,
httpd_status_codes_301_min=0,
httpd_status_codes_301_stddev=0,
httpd_status_codes_301_sum=0,
httpd_status_codes_304_current=0,
httpd_status_codes_304_max=0,
httpd_status_codes_304_mean=0,
httpd_status_codes_304_min=0,
httpd_status_codes_304_stddev=0,
httpd_status_codes_304_sum=0,
httpd_status_codes_400_current=0,
httpd_status_codes_400_max=0,
httpd_status_codes_400_mean=0,
httpd_status_codes_400_min=0,
httpd_status_codes_400_stddev=0,
httpd_status_codes_400_sum=0,
httpd_status_codes_401_current=0,
httpd_status_codes_401_max=0,
httpd_status_codes_401_mean=0,
httpd_status_codes_401_min=0,
httpd_status_codes_401_stddev=0,
httpd_status_codes_401_sum=0,
httpd_status_codes_403_current=0,
httpd_status_codes_403_max=0,
httpd_status_codes_403_mean=0,
httpd_status_codes_403_min=0,
httpd_status_codes_403_stddev=0,
httpd_status_codes_403_sum=0,
httpd_status_codes_404_current=0,
httpd_status_codes_404_max=0,
httpd_status_codes_404_mean=0,
httpd_status_codes_404_min=0,
httpd_status_codes_404_stddev=0,
httpd_status_codes_404_sum=0,
httpd_status_codes_405_current=0,
httpd_status_codes_405_max=0,
httpd_status_codes_405_mean=0,
httpd_status_codes_405_min=0,
httpd_status_codes_405_stddev=0,
httpd_status_codes_405_sum=0,
httpd_status_codes_409_current=0,
httpd_status_codes_409_max=0,
httpd_status_codes_409_mean=0,
httpd_status_codes_409_min=0,
httpd_status_codes_409_stddev=0,
httpd_status_codes_409_sum=0,
httpd_status_codes_412_current=0,
httpd_status_codes_412_max=0,
httpd_status_codes_412_mean=0,
httpd_status_codes_412_min=0,
httpd_status_codes_412_stddev=0,
httpd_status_codes_412_sum=0,
httpd_status_codes_500_current=0,
httpd_status_codes_500_max=0,
httpd_status_codes_500_mean=0,
httpd_status_codes_500_min=0,
httpd_status_codes_500_stddev=0,
httpd_status_codes_500_sum=0,
httpd_temporary_view_reads_current=0,
httpd_temporary_view_reads_max=0,
httpd_temporary_view_reads_mean=0,
httpd_temporary_view_reads_min=0,
httpd_temporary_view_reads_stddev=0,
httpd_temporary_view_reads_sum=0,
httpd_view_reads_current=0,
httpd_view_reads_max=0,
httpd_view_reads_mean=0,
httpd_view_reads_min=0,
httpd_view_reads_stddev=0,
httpd_view_reads_sum=0 1454692257621938169
```

View File

@ -0,0 +1,205 @@
package couchdb
import (
"encoding/json"
"errors"
"fmt"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/plugins/inputs"
"net/http"
"reflect"
"strings"
"sync"
)
// Schema:
type metaData struct {
Description string `json:"description"`
Current float64 `json:"current"`
Sum float64 `json:"sum"`
Mean float64 `json:"mean"`
Stddev float64 `json:"stddev"`
Min float64 `json:"min"`
Max float64 `json:"max"`
}
type Stats struct {
Couchdb struct {
AuthCacheMisses metaData `json:"auth_cache_misses"`
DatabaseWrites metaData `json:"database_writes"`
OpenDatabases metaData `json:"open_databases"`
AuthCacheHits metaData `json:"auth_cache_hits"`
RequestTime metaData `json:"request_time"`
DatabaseReads metaData `json:"database_reads"`
OpenOsFiles metaData `json:"open_os_files"`
} `json:"couchdb"`
HttpdRequestMethods struct {
Put metaData `json:"PUT"`
Get metaData `json:"GET"`
Copy metaData `json:"COPY"`
Delete metaData `json:"DELETE"`
Post metaData `json:"POST"`
Head metaData `json:"HEAD"`
} `json:"httpd_request_methods"`
HttpdStatusCodes struct {
Status200 metaData `json:"200"`
Status201 metaData `json:"201"`
Status202 metaData `json:"202"`
Status301 metaData `json:"301"`
Status304 metaData `json:"304"`
Status400 metaData `json:"400"`
Status401 metaData `json:"401"`
Status403 metaData `json:"403"`
Status404 metaData `json:"404"`
Status405 metaData `json:"405"`
Status409 metaData `json:"409"`
Status412 metaData `json:"412"`
Status500 metaData `json:"500"`
} `json:"httpd_status_codes"`
Httpd struct {
ClientsRequestingChanges metaData `json:"clients_requesting_changes"`
TemporaryViewReads metaData `json:"temporary_view_reads"`
Requests metaData `json:"requests"`
BulkRequests metaData `json:"bulk_requests"`
ViewReads metaData `json:"view_reads"`
} `json:"httpd"`
}
type CouchDB struct {
HOSTs []string `toml:"hosts"`
}
func (*CouchDB) Description() string {
return "Read CouchDB Stats from one or more servers"
}
func (*CouchDB) SampleConfig() string {
return `
### Works with CouchDB stats endpoints out of the box
### Multiple HOSTs from which to read CouchDB stats:
hosts = ["http://localhost:8086/_stats"]
`
}
func (c *CouchDB) Gather(accumulator telegraf.Accumulator) error {
errorChannel := make(chan error, len(c.HOSTs))
var wg sync.WaitGroup
for _, u := range c.HOSTs {
wg.Add(1)
go func(host string) {
defer wg.Done()
if err := c.fetchAndInsertData(accumulator, host); err != nil {
errorChannel <- fmt.Errorf("[host=%s]: %s", host, err)
}
}(u)
}
wg.Wait()
close(errorChannel)
// If there weren't any errors, we can return nil now.
if len(errorChannel) == 0 {
return nil
}
// There were errors, so join them all together as one big error.
errorStrings := make([]string, 0, len(errorChannel))
for err := range errorChannel {
errorStrings = append(errorStrings, err.Error())
}
return errors.New(strings.Join(errorStrings, "\n"))
}
func (c *CouchDB) fetchAndInsertData(accumulator telegraf.Accumulator, host string) error {
response, error := http.Get(host)
if error != nil {
return error
}
defer response.Body.Close()
var stats Stats
decoder := json.NewDecoder(response.Body)
decoder.Decode(&stats)
fields := map[string]interface{}{}
// CouchDB meta stats:
c.MapCopy(fields, c.generateFields("couchdb_auth_cache_misses", stats.Couchdb.AuthCacheMisses))
c.MapCopy(fields, c.generateFields("couchdb_database_writes", stats.Couchdb.DatabaseWrites))
c.MapCopy(fields, c.generateFields("couchdb_open_databases", stats.Couchdb.OpenDatabases))
c.MapCopy(fields, c.generateFields("couchdb_auth_cache_hits", stats.Couchdb.AuthCacheHits))
c.MapCopy(fields, c.generateFields("couchdb_request_time", stats.Couchdb.RequestTime))
c.MapCopy(fields, c.generateFields("couchdb_database_reads", stats.Couchdb.DatabaseReads))
c.MapCopy(fields, c.generateFields("couchdb_open_os_files", stats.Couchdb.OpenOsFiles))
// http request methods stats:
c.MapCopy(fields, c.generateFields("httpd_request_methods_put", stats.HttpdRequestMethods.Put))
c.MapCopy(fields, c.generateFields("httpd_request_methods_get", stats.HttpdRequestMethods.Get))
c.MapCopy(fields, c.generateFields("httpd_request_methods_copy", stats.HttpdRequestMethods.Copy))
c.MapCopy(fields, c.generateFields("httpd_request_methods_delete", stats.HttpdRequestMethods.Delete))
c.MapCopy(fields, c.generateFields("httpd_request_methods_post", stats.HttpdRequestMethods.Post))
c.MapCopy(fields, c.generateFields("httpd_request_methods_head", stats.HttpdRequestMethods.Head))
// status code stats:
c.MapCopy(fields, c.generateFields("httpd_status_codes_200", stats.HttpdStatusCodes.Status200))
c.MapCopy(fields, c.generateFields("httpd_status_codes_201", stats.HttpdStatusCodes.Status201))
c.MapCopy(fields, c.generateFields("httpd_status_codes_202", stats.HttpdStatusCodes.Status202))
c.MapCopy(fields, c.generateFields("httpd_status_codes_301", stats.HttpdStatusCodes.Status301))
c.MapCopy(fields, c.generateFields("httpd_status_codes_304", stats.HttpdStatusCodes.Status304))
c.MapCopy(fields, c.generateFields("httpd_status_codes_400", stats.HttpdStatusCodes.Status400))
c.MapCopy(fields, c.generateFields("httpd_status_codes_401", stats.HttpdStatusCodes.Status401))
c.MapCopy(fields, c.generateFields("httpd_status_codes_403", stats.HttpdStatusCodes.Status403))
c.MapCopy(fields, c.generateFields("httpd_status_codes_404", stats.HttpdStatusCodes.Status404))
c.MapCopy(fields, c.generateFields("httpd_status_codes_405", stats.HttpdStatusCodes.Status405))
c.MapCopy(fields, c.generateFields("httpd_status_codes_409", stats.HttpdStatusCodes.Status409))
c.MapCopy(fields, c.generateFields("httpd_status_codes_412", stats.HttpdStatusCodes.Status412))
c.MapCopy(fields, c.generateFields("httpd_status_codes_500", stats.HttpdStatusCodes.Status500))
// httpd stats:
c.MapCopy(fields, c.generateFields("httpd_clients_requesting_changes", stats.Httpd.ClientsRequestingChanges))
c.MapCopy(fields, c.generateFields("httpd_temporary_view_reads", stats.Httpd.TemporaryViewReads))
c.MapCopy(fields, c.generateFields("httpd_requests", stats.Httpd.Requests))
c.MapCopy(fields, c.generateFields("httpd_bulk_requests", stats.Httpd.BulkRequests))
c.MapCopy(fields, c.generateFields("httpd_view_reads", stats.Httpd.ViewReads))
tags := map[string]string{
"server": host,
}
accumulator.AddFields("couchdb", fields, tags)
return nil
}
func (*CouchDB) MapCopy(dst, src interface{}) {
dv, sv := reflect.ValueOf(dst), reflect.ValueOf(src)
for _, k := range sv.MapKeys() {
dv.SetMapIndex(k, sv.MapIndex(k))
}
}
func (*CouchDB) safeCheck(value interface{}) interface{} {
if value == nil {
return 0.0
}
return value
}
func (c *CouchDB) generateFields(prefix string, obj metaData) map[string]interface{} {
fields := map[string]interface{}{
prefix + "_current": c.safeCheck(obj.Current),
prefix + "_sum": c.safeCheck(obj.Sum),
prefix + "_mean": c.safeCheck(obj.Mean),
prefix + "_stddev": c.safeCheck(obj.Stddev),
prefix + "_min": c.safeCheck(obj.Min),
prefix + "_max": c.safeCheck(obj.Max),
}
return fields
}
func init() {
inputs.Add("couchdb", func() telegraf.Input {
return &CouchDB{}
})
}

View File

@ -0,0 +1,320 @@
package couchdb_test
import (
"github.com/influxdata/telegraf/plugins/inputs/couchdb"
"github.com/influxdata/telegraf/testutil"
"github.com/stretchr/testify/require"
"net/http"
"net/http/httptest"
"testing"
)
func TestBasic(t *testing.T) {
js := `
{
"couchdb": {
"auth_cache_misses": {
"description": "number of authentication cache misses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"database_writes": {
"description": "number of times a database was changed",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"open_databases": {
"description": "number of open databases",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"auth_cache_hits": {
"description": "number of authentication cache hits",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"request_time": {
"description": "length of a request inside CouchDB without MochiWeb",
"current": 18.0,
"sum": 18.0,
"mean": 18.0,
"stddev": null,
"min": 18.0,
"max": 18.0
},
"database_reads": {
"description": "number of times a document was read from a database",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"open_os_files": {
"description": "number of file descriptors CouchDB has open",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
}
},
"httpd_request_methods": {
"PUT": {
"description": "number of HTTP PUT requests",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"GET": {
"description": "number of HTTP GET requests",
"current": 2.0,
"sum": 2.0,
"mean": 0.25,
"stddev": 0.70699999999999996181,
"min": 0,
"max": 2
},
"COPY": {
"description": "number of HTTP COPY requests",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"DELETE": {
"description": "number of HTTP DELETE requests",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"POST": {
"description": "number of HTTP POST requests",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"HEAD": {
"description": "number of HTTP HEAD requests",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
}
},
"httpd_status_codes": {
"403": {
"description": "number of HTTP 403 Forbidden responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"202": {
"description": "number of HTTP 202 Accepted responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"401": {
"description": "number of HTTP 401 Unauthorized responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"409": {
"description": "number of HTTP 409 Conflict responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"200": {
"description": "number of HTTP 200 OK responses",
"current": 1.0,
"sum": 1.0,
"mean": 0.125,
"stddev": 0.35399999999999998135,
"min": 0,
"max": 1
},
"405": {
"description": "number of HTTP 405 Method Not Allowed responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"400": {
"description": "number of HTTP 400 Bad Request responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"201": {
"description": "number of HTTP 201 Created responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"404": {
"description": "number of HTTP 404 Not Found responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"500": {
"description": "number of HTTP 500 Internal Server Error responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"412": {
"description": "number of HTTP 412 Precondition Failed responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"301": {
"description": "number of HTTP 301 Moved Permanently responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"304": {
"description": "number of HTTP 304 Not Modified responses",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
}
},
"httpd": {
"clients_requesting_changes": {
"description": "number of clients for continuous _changes",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"temporary_view_reads": {
"description": "number of temporary view reads",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"requests": {
"description": "number of HTTP requests",
"current": 2.0,
"sum": 2.0,
"mean": 0.25,
"stddev": 0.70699999999999996181,
"min": 0,
"max": 2
},
"bulk_requests": {
"description": "number of bulk requests",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
},
"view_reads": {
"description": "number of view reads",
"current": null,
"sum": null,
"mean": null,
"stddev": null,
"min": null,
"max": null
}
}
}
`
fakeServer := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.URL.Path == "/_stats" {
_, _ = w.Write([]byte(js))
} else {
w.WriteHeader(http.StatusNotFound)
}
}))
defer fakeServer.Close()
plugin := &couchdb.CouchDB{
HOSTs: []string{fakeServer.URL + "/_stats"},
}
var acc testutil.Accumulator
require.NoError(t, plugin.Gather(&acc))
}

View File

@ -22,11 +22,11 @@ type Disque struct {
}
var sampleConfig = `
# An array of URI to gather stats about. Specify an ip or hostname
# with optional port and password. ie disque://localhost, disque://10.10.3.33:18832,
# 10.0.0.1:10000, etc.
#
# If no servers are specified, then localhost is used as the host.
### An array of URI to gather stats about. Specify an ip or hostname
### with optional port and password. ie disque://localhost, disque://10.10.3.33:18832,
### 10.0.0.1:10000, etc.
### If no servers are specified, then localhost is used as the host.
servers = ["localhost"]
`

View File

@ -21,11 +21,11 @@ type Docker struct {
}
var sampleConfig = `
# Docker Endpoint
# To use TCP, set endpoint = "tcp://[ip]:[port]"
# To use environment variables (ie, docker-machine), set endpoint = "ENV"
### Docker Endpoint
### To use TCP, set endpoint = "tcp://[ip]:[port]"
### To use environment variables (ie, docker-machine), set endpoint = "ENV"
endpoint = "unix:///var/run/docker.sock"
# Only collect metrics for these containers, collect all if empty
### Only collect metrics for these containers, collect all if empty
container_names = []
`

View File

@ -56,10 +56,13 @@ domains. You can read Dovecot's documentation
### Example Output:
```
telegraf -config telegraf.cfg -input-filter dovecot -test
>>>>>>> upstream/master
* Plugin: dovecot, Collection 1
> dovecot,domain=xxxxx.it,server=dovecot--1.mail.sys clock_time=12105746411632.5,disk_input=115285225472i,disk_output=4885067755520i,invol_cs=169701886i,last_update="2016-02-09 08:49:47.000014113 +0100 CET",mail_cache_hits=441828i,mail_lookup_attr=0i,mail_lookup_path=25323i,mail_read_bytes=241188145i,mail_read_count=11719i,maj_faults=3168i,min_faults=321438988i,num_cmds=51635i,num_connected_sessions=2i,num_logins=17149i,read_bytes=7939026951110i,read_count=3716991752i,reset_timestamp="2016-01-28 09:34:36 +0100 CET",sys_cpu=222595.288,user_cpu=267468.08,vol_cs=3288715920i,write_bytes=4483648967059i,write_count=1640646952i 1455004219924838345
> dovecot,domain=yyyyy.com,server=dovecot-1.mail.sys clock_time=6650794455331782,disk_input=61957695569920i,disk_output=2638244004487168i,invol_cs=2004805041i,last_update="2016-02-09 08:49:49.000251296 +0100 CET",mail_cache_hits=2499112513i,mail_lookup_attr=506730i,mail_lookup_path=39128227i,mail_read_bytes=1076496874501i,mail_read_count=32615262i,maj_faults=1643304i,min_faults=4216116325i,num_cmds=85785559i,num_connected_sessions=1177i,num_logins=11658255i,read_bytes=4289150974554145i,read_count=1112000703i,reset_timestamp="2016-01-28 09:31:26 +0100 CET",sys_cpu=121125923.032,user_cpu=145561336.428,vol_cs=205451885i,write_bytes=2420130526835796i,write_count=2991367252i 1455004219925152529
> dovecot,domain=xxxxx.it,server=dovecot-2.mail.sys clock_time=10710826586999.143,disk_input=79792410624i,disk_output=4496066158592i,invol_cs=150426876i,last_update="2016-02-09 08:48:19.000209134 +0100 CET",mail_cache_hits=5480869i,mail_lookup_attr=0i,mail_lookup_path=122563i,mail_read_bytes=340746273i,mail_read_count=44275i,maj_faults=1722i,min_faults=288071875i,num_cmds=50098i,num_connected_sessions=0i,num_logins=16389i,read_bytes=7259551999517i,read_count=3396625369i,reset_timestamp="2016-01-28 09:31:29 +0100 CET",sys_cpu=200762.792,user_cpu=242477.664,vol_cs=2996657358i,write_bytes=4133381575263i,write_count=1497242759i 1455004219924888283
> dovecot,domain=yyyyy.com,server=dovecot-2.mail.sys clock_time=6522131245483702,disk_input=48259150004224i,disk_output=2754333359087616i,invol_cs=2294595260i,last_update="2016-02-09 08:49:49.000251919 +0100 CET",mail_cache_hits=2139113611i,mail_lookup_attr=520276i,mail_lookup_path=37940318i,mail_read_bytes=1088002215022i,mail_read_count=31350271i,maj_faults=994420i,min_faults=1486260543i,num_cmds=40414997i,num_connected_sessions=978i,num_logins=11259672i,read_bytes=4445546612487315i,read_count=1763534543i,reset_timestamp="2016-01-28 09:31:24 +0100 CET",sys_cpu=123655962.668,user_cpu=149259327.032,vol_cs=4215130546i,write_bytes=2531186030222761i,write_count=2186579650i 1455004219925398372
```

View File

@ -24,13 +24,13 @@ func (d *Dovecot) Description() string {
}
var sampleConfig = `
# specify dovecot servers via an address:port list
# e.g.
# localhost:24242
#
# If no servers are specified, then localhost is used as the host.
### specify dovecot servers via an address:port list
### e.g.
### localhost:24242
###
### If no servers are specified, then localhost is used as the host.
servers = ["localhost:24242"]
# Only collect metrics for these domains, collect all if empty
### Only collect metrics for these domains, collect all if empty
domains = []
`

View File

@ -10,8 +10,8 @@ import (
"time"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/internal"
"github.com/influxdata/telegraf/plugins/inputs"
jsonparser "github.com/influxdata/telegraf/plugins/parsers/json"
)
const statsPath = "/_nodes/stats"
@ -59,14 +59,14 @@ type indexHealth struct {
}
const sampleConfig = `
# specify a list of one or more Elasticsearch servers
### specify a list of one or more Elasticsearch servers
servers = ["http://localhost:9200"]
# set local to false when you want to read the indices stats from all nodes
# within the cluster
### set local to false when you want to read the indices stats from all nodes
### within the cluster
local = true
# set cluster_health to true when you want to also obtain cluster level stats
### set cluster_health to true when you want to also obtain cluster level stats
cluster_health = false
`
@ -168,7 +168,7 @@ func (e *Elasticsearch) gatherNodeStats(url string, acc telegraf.Accumulator) er
now := time.Now()
for p, s := range stats {
f := internal.JSONFlattener{}
f := jsonparser.JSONFlattener{}
err := f.FlattenJSON("", s)
if err != nil {
return err

View File

@ -1,7 +1,23 @@
# Exec Input Plugin
The exec plugin can execute arbitrary commands which output JSON or
InfluxDB [line-protocol](https://docs.influxdata.com/influxdb/v0.9/write_protocols/line/).
The exec plugin can execute arbitrary commands which output:
* JSON
* InfluxDB [line-protocol](https://docs.influxdata.com/influxdb/v0.9/write_protocols/line/)
* Graphite [graphite-protocol](http://graphite.readthedocs.org/en/latest/feeding-carbon.html)
> Graphite understands messages with this format:
> ```
metric_path value timestamp\n
```
> __metric_path__ is the metric namespace that you want to populate.
> __value__ is the value that you want to assign to the metric at this time.
> __timestamp__ is the unix epoch time.
If using JSON, only numeric values are parsed and turned into floats. Booleans
and strings will be ignored.
@ -11,21 +27,40 @@ and strings will be ignored.
```
# Read flattened metrics from one or more commands that output JSON to stdout
[[inputs.exec]]
# the command to run
command = "/usr/bin/mycollector --foo=bar"
# Shell/commands array
commands = ["/tmp/test.sh", "/tmp/test2.sh"]
# Data format to consume. This can be "json" or "influx" (line-protocol)
# Data format to consume. This can be "json", "influx" or "graphite" (line-protocol)
# NOTE json only reads numerical measurements, strings and booleans are ignored.
data_format = "json"
# measurement name suffix (for separating different commands)
name_suffix = "_mycollector"
### Below configuration will be used for data_format = "graphite", can be ignored for other data_format
### If matching multiple measurement files, this string will be used to join the matched values.
#separator = "."
### Each template line requires a template pattern. It can have an optional
### filter before the template and separated by spaces. It can also have optional extra
### tags following the template. Multiple tags should be separated by commas and no spaces
### similar to the line protocol format. The can be only one default template.
### Templates support below format:
### 1. filter + template
### 2. filter + template + extra tag
### 3. filter + template with field key
### 4. default template
#templates = [
# "*.app env.service.resource.measurement",
# "stats.* .host.measurement* region=us-west,agent=sensu",
# "stats2.* .host.measurement.field",
# "measurement*"
#]
```
Other options for modifying the measurement names are:
```
name_override = "measurement_name"
name_prefix = "prefix_"
```
@ -57,8 +92,11 @@ Now let's say we have the following configuration:
```
[[inputs.exec]]
# the command to run
command = "/usr/bin/line_protocol_collector"
# Shell/commands array
# compatible with old version
# we can still use the old command configuration
# command = "/usr/bin/line_protocol_collector"
commands = ["/usr/bin/line_protocol_collector","/tmp/test2.sh"]
# Data format to consume. This can be "json" or "influx" (line-protocol)
# NOTE json only reads numerical measurements, strings and booleans are ignored.
@ -80,3 +118,63 @@ cpu,cpu=cpu6,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
You will get data in InfluxDB exactly as it is defined above,
tags are cpu=cpuN, host=foo, and datacenter=us-east with fields usage_idle
and usage_busy. They will receive a timestamp at collection time.
### Example 3
We can also change the data_format to "graphite" to use the metrics collecting scripts such as (compatible with graphite):
* Nagios [Mertics Plugins] (https://exchange.nagios.org/directory/Plugins)
* Sensu [Mertics Plugins] (https://github.com/sensu-plugins)
#### Configuration
```
# Read flattened metrics from one or more commands that output JSON to stdout
[[inputs.exec]]
# Shell/commands array
commands = ["/tmp/test.sh","/tmp/test2.sh"]
# Data format to consume. This can be "json", "influx" or "graphite" (line-protocol)
# NOTE json only reads numerical measurements, strings and booleans are ignored.
data_format = "graphite"
# measurement name suffix (for separating different commands)
name_suffix = "_mycollector"
### Below configuration will be used for data_format = "graphite", can be ignored for other data_format
### If matching multiple measurement files, this string will be used to join the matched values.
separator = "."
### Each template line requires a template pattern. It can have an optional
### filter before the template and separated by spaces. It can also have optional extra
### tags following the template. Multiple tags should be separated by commas and no spaces
### similar to the line protocol format. The can be only one default template.
### Templates support below format:
### 1. filter + template
### 2. filter + template + extra tag
### 3. filter + template with field key
### 4. default template
templates = [
"*.app env.service.resource.measurement",
"stats.* .host.measurement* region=us-west,agent=sensu",
"stats2.* .host.measurement.field",
"measurement*"
]
```
And test.sh/test2.sh will output:
```
sensu.metric.net.server0.eth0.rx_packets 461295119435 1444234982
sensu.metric.net.server0.eth0.tx_bytes 1093086493388480 1444234982
sensu.metric.net.server0.eth0.rx_bytes 1015633926034834 1444234982
sensu.metric.net.server0.eth0.tx_errors 0 1444234982
sensu.metric.net.server0.eth0.rx_errors 0 1444234982
sensu.metric.net.server0.eth0.tx_dropped 0 1444234982
sensu.metric.net.server0.eth0.rx_dropped 0 1444234982
```
The templates configuration will be used to parse the graphite metrics to support influxdb/opentsdb tagging store engines.
More detail information about templates, please refer to [The graphite Input] (https://github.com/influxdata/influxdb/blob/master/services/graphite/README.md)

View File

@ -2,62 +2,90 @@ package exec
import (
"bytes"
"encoding/json"
"fmt"
"os/exec"
"time"
"sync"
"github.com/gonuts/go-shellquote"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/internal"
"github.com/influxdata/telegraf/plugins/inputs"
"github.com/influxdata/telegraf/plugins/parsers"
)
const sampleConfig = `
# the command to run
command = "/usr/bin/mycollector --foo=bar"
### Commands array
commands = ["/tmp/test.sh", "/usr/bin/mycollector --foo=bar"]
# Data format to consume. This can be "json" or "influx" (line-protocol)
# NOTE json only reads numerical measurements, strings and booleans are ignored.
data_format = "json"
# measurement name suffix (for separating different commands)
### measurement name suffix (for separating different commands)
name_suffix = "_mycollector"
### Data format to consume. This can be "json", "influx" or "graphite"
### Each data format has it's own unique set of configuration options, read
### more about them here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md
data_format = "influx"
`
type Exec struct {
Commands []string
Command string
DataFormat string
parser parsers.Parser
wg sync.WaitGroup
runner Runner
errChan chan error
}
func NewExec() *Exec {
return &Exec{
runner: CommandRunner{},
}
}
type Runner interface {
Run(*Exec) ([]byte, error)
Run(*Exec, string) ([]byte, error)
}
type CommandRunner struct{}
func (c CommandRunner) Run(e *Exec) ([]byte, error) {
split_cmd, err := shellquote.Split(e.Command)
func (c CommandRunner) Run(e *Exec, command string) ([]byte, error) {
split_cmd, err := shellquote.Split(command)
if err != nil || len(split_cmd) == 0 {
return nil, fmt.Errorf("exec: unable to parse command, %s", err)
}
cmd := exec.Command(split_cmd[0], split_cmd[1:]...)
var out bytes.Buffer
cmd.Stdout = &out
if err := cmd.Run(); err != nil {
return nil, fmt.Errorf("exec: %s for command '%s'", err, e.Command)
return nil, fmt.Errorf("exec: %s for command '%s'", err, command)
}
return out.Bytes(), nil
}
func NewExec() *Exec {
return &Exec{runner: CommandRunner{}}
func (e *Exec) ProcessCommand(command string, acc telegraf.Accumulator) {
defer e.wg.Done()
out, err := e.runner.Run(e, command)
if err != nil {
e.errChan <- err
return
}
metrics, err := e.parser.Parse(out)
if err != nil {
e.errChan <- err
} else {
for _, metric := range metrics {
acc.AddFields(metric.Name(), metric.Fields(), metric.Tags(), metric.Time())
}
}
}
func (e *Exec) SampleConfig() string {
@ -65,42 +93,37 @@ func (e *Exec) SampleConfig() string {
}
func (e *Exec) Description() string {
return "Read flattened metrics from one or more commands that output JSON to stdout"
return "Read metrics from one or more commands that can output to stdout"
}
func (e *Exec) SetParser(parser parsers.Parser) {
e.parser = parser
}
func (e *Exec) Gather(acc telegraf.Accumulator) error {
out, err := e.runner.Run(e)
if err != nil {
return err
// Legacy single command support
if e.Command != "" {
e.Commands = append(e.Commands, e.Command)
e.Command = ""
}
switch e.DataFormat {
case "", "json":
var jsonOut interface{}
err = json.Unmarshal(out, &jsonOut)
if err != nil {
return fmt.Errorf("exec: unable to parse output of '%s' as JSON, %s",
e.Command, err)
}
e.errChan = make(chan error, len(e.Commands))
f := internal.JSONFlattener{}
err = f.FlattenJSON("", jsonOut)
if err != nil {
return err
e.wg.Add(len(e.Commands))
for _, command := range e.Commands {
go e.ProcessCommand(command, acc)
}
acc.AddFields("exec", f.Fields, nil)
case "influx":
now := time.Now()
metrics, err := telegraf.ParseMetrics(out)
for _, metric := range metrics {
acc.AddFields(metric.Name(), metric.Fields(), metric.Tags(), now)
}
return err
e.wg.Wait()
select {
default:
return fmt.Errorf("Unsupported data format: %s. Must be either json "+
"or influx.", e.DataFormat)
}
close(e.errChan)
return nil
case err := <-e.errChan:
close(e.errChan)
return err
}
}
func init() {

View File

@ -4,6 +4,8 @@ import (
"fmt"
"testing"
"github.com/influxdata/telegraf/plugins/parsers"
"github.com/influxdata/telegraf/testutil"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
@ -55,7 +57,7 @@ func newRunnerMock(out []byte, err error) Runner {
}
}
func (r runnerMock) Run(e *Exec) ([]byte, error) {
func (r runnerMock) Run(e *Exec, command string) ([]byte, error) {
if r.err != nil {
return nil, r.err
}
@ -63,9 +65,11 @@ func (r runnerMock) Run(e *Exec) ([]byte, error) {
}
func TestExec(t *testing.T) {
parser, _ := parsers.NewJSONParser("exec", []string{}, nil)
e := &Exec{
runner: newRunnerMock([]byte(validJson), nil),
Command: "testcommand arg1",
Commands: []string{"testcommand arg1"},
parser: parser,
}
var acc testutil.Accumulator
@ -87,9 +91,11 @@ func TestExec(t *testing.T) {
}
func TestExecMalformed(t *testing.T) {
parser, _ := parsers.NewJSONParser("exec", []string{}, nil)
e := &Exec{
runner: newRunnerMock([]byte(malformedJson), nil),
Command: "badcommand arg1",
Commands: []string{"badcommand arg1"},
parser: parser,
}
var acc testutil.Accumulator
@ -99,9 +105,11 @@ func TestExecMalformed(t *testing.T) {
}
func TestCommandError(t *testing.T) {
parser, _ := parsers.NewJSONParser("exec", []string{}, nil)
e := &Exec{
runner: newRunnerMock(nil, fmt.Errorf("exit status code 1")),
Command: "badcommand",
Commands: []string{"badcommand"},
parser: parser,
}
var acc testutil.Accumulator
@ -111,10 +119,11 @@ func TestCommandError(t *testing.T) {
}
func TestLineProtocolParse(t *testing.T) {
parser, _ := parsers.NewInfluxParser()
e := &Exec{
runner: newRunnerMock([]byte(lineProtocol), nil),
Command: "line-protocol",
DataFormat: "influx",
Commands: []string{"line-protocol"},
parser: parser,
}
var acc testutil.Accumulator
@ -133,10 +142,11 @@ func TestLineProtocolParse(t *testing.T) {
}
func TestLineProtocolParseMultiple(t *testing.T) {
parser, _ := parsers.NewInfluxParser()
e := &Exec{
runner: newRunnerMock([]byte(lineProtocolMulti), nil),
Command: "line-protocol",
DataFormat: "influx",
Commands: []string{"line-protocol"},
parser: parser,
}
var acc testutil.Accumulator
@ -158,15 +168,3 @@ func TestLineProtocolParseMultiple(t *testing.T) {
acc.AssertContainsTaggedFields(t, "cpu", fields, tags)
}
}
func TestInvalidDataFormat(t *testing.T) {
e := &Exec{
runner: newRunnerMock([]byte(lineProtocol), nil),
Command: "bad data format",
DataFormat: "FooBar",
}
var acc testutil.Accumulator
err := e.Gather(&acc)
require.Error(t, err)
}

View File

@ -31,7 +31,7 @@ func NewGithubWebhooks() *GithubWebhooks {
func (gh *GithubWebhooks) SampleConfig() string {
return `
# Address and port to host Webhook listener on
### Address and port to host Webhook listener on
service_address = ":1618"
`
}

View File

@ -86,13 +86,13 @@ type haproxy struct {
}
var sampleConfig = `
# An array of address to gather stats about. Specify an ip on hostname
# with optional port. ie localhost, 10.10.3.33:1936, etc.
#
# If no servers are specified, then default to 127.0.0.1:1936
### An array of address to gather stats about. Specify an ip on hostname
### with optional port. ie localhost, 10.10.3.33:1936, etc.
### If no servers are specified, then default to 127.0.0.1:1936
servers = ["http://myhaproxy.com:1936", "http://anotherhaproxy.com:1936"]
# Or you can also use local socket(not work yet)
# servers = ["socket://run/haproxy/admin.sock"]
### Or you can also use local socket(not work yet)
### servers = ["socket://run/haproxy/admin.sock"]
`
func (r *haproxy) SampleConfig() string {

View File

@ -1,7 +1,6 @@
package httpjson
import (
"encoding/json"
"errors"
"fmt"
"io/ioutil"
@ -12,8 +11,8 @@ import (
"time"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/internal"
"github.com/influxdata/telegraf/plugins/inputs"
"github.com/influxdata/telegraf/plugins/parsers"
)
type HttpJson struct {
@ -47,37 +46,36 @@ func (c RealHTTPClient) MakeRequest(req *http.Request) (*http.Response, error) {
}
var sampleConfig = `
# NOTE This plugin only reads numerical measurements, strings and booleans
# will be ignored.
### NOTE This plugin only reads numerical measurements, strings and booleans
### will be ignored.
# a name for the service being polled
### a name for the service being polled
name = "webserver_stats"
# URL of each server in the service's cluster
### URL of each server in the service's cluster
servers = [
"http://localhost:9999/stats/",
"http://localhost:9998/stats/",
]
# HTTP method to use (case-sensitive)
### HTTP method to use (case-sensitive)
method = "GET"
# List of tag names to extract from top-level of JSON server response
### List of tag names to extract from top-level of JSON server response
# tag_keys = [
# "my_tag_1",
# "my_tag_2"
# ]
# HTTP parameters (all values must be strings)
### HTTP parameters (all values must be strings)
[inputs.httpjson.parameters]
event_type = "cpu_spike"
threshold = "0.75"
# HTTP Header parameters (all values must be strings)
### HTTP Header parameters (all values must be strings)
# [inputs.httpjson.headers]
# X-Auth-Token = "my-xauth-token"
# apiVersion = "v1"
`
func (h *HttpJson) SampleConfig() string {
@ -137,39 +135,34 @@ func (h *HttpJson) gatherServer(
return err
}
var jsonOut map[string]interface{}
if err = json.Unmarshal([]byte(resp), &jsonOut); err != nil {
return errors.New("Error decoding JSON response")
}
tags := map[string]string{
"server": serverURL,
}
for _, tag := range h.TagKeys {
switch v := jsonOut[tag].(type) {
case string:
tags[tag] = v
}
delete(jsonOut, tag)
}
if responseTime >= 0 {
jsonOut["response_time"] = responseTime
}
f := internal.JSONFlattener{}
err = f.FlattenJSON("", jsonOut)
if err != nil {
return err
}
var msrmnt_name string
if h.Name == "" {
msrmnt_name = "httpjson"
} else {
msrmnt_name = "httpjson_" + h.Name
}
acc.AddFields(msrmnt_name, f.Fields, tags)
tags := map[string]string{
"server": serverURL,
}
parser, err := parsers.NewJSONParser(msrmnt_name, h.TagKeys, tags)
if err != nil {
return err
}
metrics, err := parser.Parse([]byte(resp))
if err != nil {
return err
}
for _, metric := range metrics {
fields := make(map[string]interface{})
for k, v := range metric.Fields() {
fields[k] = v
}
fields["response_time"] = responseTime
acc.AddFields(metric.Name(), fields, metric.Tags())
}
return nil
}

View File

@ -22,11 +22,11 @@ func (*InfluxDB) Description() string {
func (*InfluxDB) SampleConfig() string {
return `
# Works with InfluxDB debug endpoints out of the box,
# but other services can use this format too.
# See the influxdb plugin's README for more details.
### Works with InfluxDB debug endpoints out of the box,
### but other services can use this format too.
### See the influxdb plugin's README for more details.
# Multiple URLs from which to read InfluxDB-formatted JSON
### Multiple URLs from which to read InfluxDB-formatted JSON
urls = [
"http://localhost:8086/debug/vars"
]

View File

@ -46,10 +46,10 @@ type Jolokia struct {
func (j *Jolokia) SampleConfig() string {
return `
# This is the context root used to compose the jolokia url
### This is the context root used to compose the jolokia url
context = "/jolokia/read"
# List of servers exposing jolokia read service
### List of servers exposing jolokia read service
[[inputs.jolokia.servers]]
name = "stable"
host = "192.168.103.2"
@ -57,9 +57,10 @@ func (j *Jolokia) SampleConfig() string {
# username = "myuser"
# password = "mypassword"
# List of metrics collected on above servers
# Each metric consists in a name, a jmx path and either a pass or drop slice attributes
# This collect all heap memory usage metrics
### List of metrics collected on above servers
### Each metric consists in a name, a jmx path and either
### a pass or drop slice attribute.
### This collect all heap memory usage metrics.
[[inputs.jolokia.metrics]]
name = "heap_memory_usage"
jmx = "/java.lang:type=Memory/HeapMemoryUsage"

View File

@ -1,12 +1,14 @@
package kafka_consumer
import (
"fmt"
"log"
"strings"
"sync"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/plugins/inputs"
"github.com/influxdata/telegraf/plugins/parsers"
"github.com/Shopify/sarama"
"github.com/wvanbergen/kafka/consumergroup"
@ -20,6 +22,8 @@ type Kafka struct {
PointBuffer int
Offset string
parser parsers.Parser
sync.Mutex
// channel for all incoming kafka messages
@ -36,16 +40,22 @@ type Kafka struct {
}
var sampleConfig = `
# topic(s) to consume
### topic(s) to consume
topics = ["telegraf"]
# an array of Zookeeper connection strings
### an array of Zookeeper connection strings
zookeeper_peers = ["localhost:2181"]
# the name of the consumer group
### the name of the consumer group
consumer_group = "telegraf_metrics_consumers"
# Maximum number of points to buffer between collection intervals
### Maximum number of points to buffer between collection intervals
point_buffer = 100000
# Offset (must be either "oldest" or "newest")
### Offset (must be either "oldest" or "newest")
offset = "oldest"
### Data format to consume. This can be "json", "influx" or "graphite"
### Each data format has it's own unique set of configuration options, read
### more about them here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md
data_format = "influx"
`
func (k *Kafka) SampleConfig() string {
@ -53,7 +63,11 @@ func (k *Kafka) SampleConfig() string {
}
func (k *Kafka) Description() string {
return "Read line-protocol metrics from Kafka topic(s)"
return "Read metrics from Kafka topic(s)"
}
func (k *Kafka) SetParser(parser parsers.Parser) {
k.parser = parser
}
func (k *Kafka) Start() error {
@ -96,15 +110,15 @@ func (k *Kafka) Start() error {
k.metricC = make(chan telegraf.Metric, k.PointBuffer)
// Start the kafka message reader
go k.parser()
go k.receiver()
log.Printf("Started the kafka consumer service, peers: %v, topics: %v\n",
k.ZookeeperPeers, k.Topics)
return nil
}
// parser() reads all incoming messages from the consumer, and parses them into
// receiver() reads all incoming messages from the consumer, and parses them into
// influxdb metric points.
func (k *Kafka) parser() {
func (k *Kafka) receiver() {
for {
select {
case <-k.done:
@ -112,13 +126,14 @@ func (k *Kafka) parser() {
case err := <-k.errs:
log.Printf("Kafka Consumer Error: %s\n", err.Error())
case msg := <-k.in:
metrics, err := telegraf.ParseMetrics(msg.Value)
metrics, err := k.parser.Parse(msg.Value)
if err != nil {
log.Printf("Could not parse kafka message: %s, error: %s",
string(msg.Value), err.Error())
}
for _, metric := range metrics {
fmt.Println(string(metric.Name()))
select {
case k.metricC <- metric:
continue

View File

@ -9,6 +9,8 @@ import (
"github.com/influxdata/telegraf/testutil"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
"github.com/influxdata/telegraf/plugins/parsers"
)
func TestReadsMetricsFromKafka(t *testing.T) {
@ -40,6 +42,8 @@ func TestReadsMetricsFromKafka(t *testing.T) {
PointBuffer: 100000,
Offset: "oldest",
}
p, _ := parsers.NewInfluxParser()
k.SetParser(p)
if err := k.Start(); err != nil {
t.Fatal(err.Error())
} else {

View File

@ -5,6 +5,7 @@ import (
"time"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/plugins/parsers"
"github.com/influxdata/telegraf/testutil"
"github.com/Shopify/sarama"
@ -13,6 +14,8 @@ import (
const (
testMsg = "cpu_load_short,host=server01 value=23422.0 1422568543702900257"
testMsgGraphite = "cpu.load.short.graphite 23422 1454780029"
testMsgJSON = "{\"a\": 5, \"b\": {\"c\": 6}}\n"
invalidMsg = "cpu_load_short,host=server01 1422568543702900257"
pointBuffer = 5
)
@ -39,7 +42,8 @@ func TestRunParser(t *testing.T) {
k, in := NewTestKafka()
defer close(k.done)
go k.parser()
k.parser, _ = parsers.NewInfluxParser()
go k.receiver()
in <- saramaMsg(testMsg)
time.Sleep(time.Millisecond)
@ -51,7 +55,8 @@ func TestRunParserInvalidMsg(t *testing.T) {
k, in := NewTestKafka()
defer close(k.done)
go k.parser()
k.parser, _ = parsers.NewInfluxParser()
go k.receiver()
in <- saramaMsg(invalidMsg)
time.Sleep(time.Millisecond)
@ -63,7 +68,8 @@ func TestRunParserRespectsBuffer(t *testing.T) {
k, in := NewTestKafka()
defer close(k.done)
go k.parser()
k.parser, _ = parsers.NewInfluxParser()
go k.receiver()
for i := 0; i < pointBuffer+1; i++ {
in <- saramaMsg(testMsg)
}
@ -77,7 +83,8 @@ func TestRunParserAndGather(t *testing.T) {
k, in := NewTestKafka()
defer close(k.done)
go k.parser()
k.parser, _ = parsers.NewInfluxParser()
go k.receiver()
in <- saramaMsg(testMsg)
time.Sleep(time.Millisecond)
@ -89,6 +96,45 @@ func TestRunParserAndGather(t *testing.T) {
map[string]interface{}{"value": float64(23422)})
}
// Test that the parser parses kafka messages into points
func TestRunParserAndGatherGraphite(t *testing.T) {
k, in := NewTestKafka()
defer close(k.done)
k.parser, _ = parsers.NewGraphiteParser("_", []string{}, nil)
go k.receiver()
in <- saramaMsg(testMsgGraphite)
time.Sleep(time.Millisecond)
acc := testutil.Accumulator{}
k.Gather(&acc)
assert.Equal(t, len(acc.Metrics), 1)
acc.AssertContainsFields(t, "cpu_load_short_graphite",
map[string]interface{}{"value": float64(23422)})
}
// Test that the parser parses kafka messages into points
func TestRunParserAndGatherJSON(t *testing.T) {
k, in := NewTestKafka()
defer close(k.done)
k.parser, _ = parsers.NewJSONParser("kafka_json_test", []string{}, nil)
go k.receiver()
in <- saramaMsg(testMsgJSON)
time.Sleep(time.Millisecond)
acc := testutil.Accumulator{}
k.Gather(&acc)
assert.Equal(t, len(acc.Metrics), 1)
acc.AssertContainsFields(t, "kafka_json_test",
map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
})
}
func saramaMsg(val string) *sarama.ConsumerMessage {
return &sarama.ConsumerMessage{
Key: nil,

View File

@ -132,10 +132,8 @@ var serverTypeMapping = map[string]ServerType{
}
var sampleConfig = `
# An array of URI to gather stats about LeoFS.
# Specify an ip or hostname with port. ie 127.0.0.1:4020
#
# If no servers are specified, then 127.0.0.1 is used as the host and 4020 as the port.
### An array of URI to gather stats about LeoFS.
### Specify an ip or hostname with port. ie 127.0.0.1:4020
servers = ["127.0.0.1:4021"]
`

View File

@ -29,10 +29,13 @@ type Lustre2 struct {
}
var sampleConfig = `
# An array of /proc globs to search for Lustre stats
# If not specified, the default will work on Lustre 2.5.x
#
# ost_procfiles = ["/proc/fs/lustre/obdfilter/*/stats", "/proc/fs/lustre/osd-ldiskfs/*/stats"]
### An array of /proc globs to search for Lustre stats
### If not specified, the default will work on Lustre 2.5.x
###
# ost_procfiles = [
# "/proc/fs/lustre/obdfilter/*/stats",
# "/proc/fs/lustre/osd-ldiskfs/*/stats"
# ]
# mds_procfiles = ["/proc/fs/lustre/mdt/*/md_stats"]
`

View File

@ -17,13 +17,13 @@ type MailChimp struct {
}
var sampleConfig = `
# MailChimp API key
# get from https://admin.mailchimp.com/account/api/
### MailChimp API key
### get from https://admin.mailchimp.com/account/api/
api_key = "" # required
# Reports for campaigns sent more than days_old ago will not be collected.
# 0 means collect all.
### Reports for campaigns sent more than days_old ago will not be collected.
### 0 means collect all.
days_old = 0
# Campaign ID to get, if empty gets all campaigns, this option overrides days_old
### Campaign ID to get, if empty gets all campaigns, this option overrides days_old
# campaign_id = ""
`

View File

@ -19,10 +19,8 @@ type Memcached struct {
}
var sampleConfig = `
# An array of address to gather stats about. Specify an ip on hostname
# with optional port. ie localhost, 10.0.0.1:11211, etc.
#
# If no servers are specified, then localhost is used as the host.
### An array of address to gather stats about. Specify an ip on hostname
### with optional port. ie localhost, 10.0.0.1:11211, etc.
servers = ["localhost:11211"]
# unix_sockets = ["/var/run/memcached.sock"]
`

View File

@ -26,11 +26,11 @@ type Ssl struct {
}
var sampleConfig = `
# An array of URI to gather stats about. Specify an ip or hostname
# with optional port add password. ie mongodb://user:auth_key@10.10.3.30:27017,
# mongodb://10.10.3.33:18832, 10.0.0.1:10000, etc.
#
# If no servers are specified, then 127.0.0.1 is used as the host and 27107 as the port.
### An array of URI to gather stats about. Specify an ip or hostname
### with optional port add password. ie,
### mongodb://user:auth_key@10.10.3.30:27017,
### mongodb://10.10.3.33:18832,
### 10.0.0.1:10000, etc.
servers = ["127.0.0.1:27017"]
`

View File

@ -15,14 +15,14 @@ type Mysql struct {
}
var sampleConfig = `
# specify servers via a url matching:
# [username[:password]@][protocol[(address)]]/[?tls=[true|false|skip-verify]]
# see https://github.com/go-sql-driver/mysql#dsn-data-source-name
# e.g.
# root:passwd@tcp(127.0.0.1:3306)/?tls=false
# root@tcp(127.0.0.1:3306)/?tls=false
#
# If no servers are specified, then localhost is used as the host.
### specify servers via a url matching:
### [username[:password]@][protocol[(address)]]/[?tls=[true|false|skip-verify]]
### see https://github.com/go-sql-driver/mysql#dsn-data-source-name
### e.g.
### root:passwd@tcp(127.0.0.1:3306)/?tls=false
### root@tcp(127.0.0.1:3306)/?tls=false
###
### If no servers are specified, then localhost is used as the host.
servers = ["tcp(127.0.0.1:3306)/"]
`

View File

@ -20,7 +20,7 @@ type Nginx struct {
}
var sampleConfig = `
# An array of Nginx stub_status URI to gather stats.
### An array of Nginx stub_status URI to gather stats.
urls = ["http://localhost/status"]
`

View File

@ -41,7 +41,7 @@ type NSQ struct {
}
var sampleConfig = `
# An array of NSQD HTTP API endpoints
### An array of NSQD HTTP API endpoints
endpoints = ["http://localhost:4151"]
`

View File

@ -126,15 +126,14 @@ func (p *process) getUptime() int64 {
}
var sampleConfig = `
# Path of passenger-status.
#
# Plugin gather metric via parsing XML output of passenger-status
# More information about the tool:
# https://www.phusionpassenger.com/library/admin/apache/overall_status_report.html
#
#
# If no path is specified, then the plugin simply execute passenger-status
# hopefully it can be found in your PATH
### Path of passenger-status.
###
### Plugin gather metric via parsing XML output of passenger-status
### More information about the tool:
### https://www.phusionpassenger.com/library/admin/apache/overall_status_report.html
###
### If no path is specified, then the plugin simply execute passenger-status
### hopefully it can be found in your PATH
command = "passenger-status -v --show=xml"
`

View File

@ -41,26 +41,25 @@ type phpfpm struct {
}
var sampleConfig = `
# An array of addresses to gather stats about. Specify an ip or hostname
# with optional port and path
#
# Plugin can be configured in three modes (either can be used):
# - http: the URL must start with http:// or https://, ie:
# "http://localhost/status"
# "http://192.168.130.1/status?full"
#
# - unixsocket: path to fpm socket, ie:
# "/var/run/php5-fpm.sock"
# or using a custom fpm status path:
# "/var/run/php5-fpm.sock:fpm-custom-status-path"
#
# - fcgi: the URL must start with fcgi:// or cgi://, and port must be present, ie:
# "fcgi://10.0.0.12:9000/status"
# "cgi://10.0.10.12:9001/status"
#
# Example of multiple gathering from local socket and remove host
# urls = ["http://192.168.1.20/status", "/tmp/fpm.sock"]
# If no servers are specified, then default to http://127.0.0.1/status
### An array of addresses to gather stats about. Specify an ip or hostname
### with optional port and path
###
### Plugin can be configured in three modes (either can be used):
### - http: the URL must start with http:// or https://, ie:
### "http://localhost/status"
### "http://192.168.130.1/status?full"
###
### - unixsocket: path to fpm socket, ie:
### "/var/run/php5-fpm.sock"
### or using a custom fpm status path:
### "/var/run/php5-fpm.sock:fpm-custom-status-path"
###
### - fcgi: the URL must start with fcgi:// or cgi://, and port must be present, ie:
### "fcgi://10.0.0.12:9000/status"
### "cgi://10.0.10.12:9001/status"
###
### Example of multiple gathering from local socket and remove host
### urls = ["http://192.168.1.20/status", "/tmp/fpm.sock"]
urls = ["http://localhost/status"]
`

View File

@ -5,6 +5,7 @@ package ping
import (
"errors"
"os/exec"
"runtime"
"strconv"
"strings"
"sync"
@ -43,18 +44,18 @@ func (_ *Ping) Description() string {
}
var sampleConfig = `
# NOTE: this plugin forks the ping command. You may need to set capabilities
# via setcap cap_net_raw+p /bin/ping
### NOTE: this plugin forks the ping command. You may need to set capabilities
### via setcap cap_net_raw+p /bin/ping
# urls to ping
### urls to ping
urls = ["www.google.com"] # required
# number of pings to send (ping -c <COUNT>)
### number of pings to send (ping -c <COUNT>)
count = 1 # required
# interval, in s, at which to ping. 0 == default (ping -i <PING_INTERVAL>)
### interval, in s, at which to ping. 0 == default (ping -i <PING_INTERVAL>)
ping_interval = 0.0
# ping timeout, in s. 0 == no timeout (ping -t <TIMEOUT>)
### ping timeout, in s. 0 == no timeout (ping -t <TIMEOUT>)
timeout = 0.0
# interface to send ping from (ping -I <INTERFACE>)
### interface to send ping from (ping -I <INTERFACE>)
interface = ""
`
@ -128,12 +129,20 @@ func hostPinger(args ...string) (string, error) {
// args returns the arguments for the 'ping' executable
func (p *Ping) args(url string) []string {
// Build the ping command args based on toml config
args := []string{"-c", strconv.Itoa(p.Count)}
args := []string{"-c", strconv.Itoa(p.Count), "-n", "-s", "16"}
if p.PingInterval > 0 {
args = append(args, "-i", strconv.FormatFloat(p.PingInterval, 'f', 1, 64))
}
if p.Timeout > 0 {
switch runtime.GOOS {
case "darwin", "freebsd":
args = append(args, "-t", strconv.FormatFloat(p.Timeout, 'f', 1, 64))
case "linux":
args = append(args, "-W", strconv.FormatFloat(p.Timeout, 'f', 1, 64))
default:
// Not sure the best option here, just assume GNU ping?
args = append(args, "-W", strconv.FormatFloat(p.Timeout, 'f', 1, 64))
}
}
if p.Interface != "" {
args = append(args, "-I", p.Interface)

View File

@ -5,6 +5,7 @@ package ping
import (
"errors"
"reflect"
"runtime"
"sort"
"testing"
@ -76,7 +77,7 @@ func TestArgs(t *testing.T) {
// Actual and Expected arg lists must be sorted for reflect.DeepEqual
actual := p.args("www.google.com")
expected := []string{"-c", "2", "www.google.com"}
expected := []string{"-c", "2", "-n", "-s", "16", "www.google.com"}
sort.Strings(actual)
sort.Strings(expected)
assert.True(t, reflect.DeepEqual(expected, actual),
@ -84,7 +85,8 @@ func TestArgs(t *testing.T) {
p.Interface = "eth0"
actual = p.args("www.google.com")
expected = []string{"-c", "2", "-I", "eth0", "www.google.com"}
expected = []string{"-c", "2", "-n", "-s", "16", "-I", "eth0",
"www.google.com"}
sort.Strings(actual)
sort.Strings(expected)
assert.True(t, reflect.DeepEqual(expected, actual),
@ -92,7 +94,15 @@ func TestArgs(t *testing.T) {
p.Timeout = 12.0
actual = p.args("www.google.com")
expected = []string{"-c", "2", "-I", "eth0", "-t", "12.0", "www.google.com"}
switch runtime.GOOS {
case "darwin", "freebsd":
expected = []string{"-c", "2", "-n", "-s", "16", "-I", "eth0", "-t",
"12.0", "www.google.com"}
default:
expected = []string{"-c", "2", "-n", "-s", "16", "-I", "eth0", "-W",
"12.0", "www.google.com"}
}
sort.Strings(actual)
sort.Strings(expected)
assert.True(t, reflect.DeepEqual(expected, actual),
@ -100,8 +110,14 @@ func TestArgs(t *testing.T) {
p.PingInterval = 1.2
actual = p.args("www.google.com")
expected = []string{"-c", "2", "-I", "eth0", "-t", "12.0", "-i", "1.2",
"www.google.com"}
switch runtime.GOOS {
case "darwin", "freebsd":
expected = []string{"-c", "2", "-n", "-s", "16", "-I", "eth0", "-t",
"12.0", "-i", "1.2", "www.google.com"}
default:
expected = []string{"-c", "2", "-n", "-s", "16", "-I", "eth0", "-W",
"12.0", "-i", "1.2", "www.google.com"}
}
sort.Strings(actual)
sort.Strings(expected)
assert.True(t, reflect.DeepEqual(expected, actual),

View File

@ -21,22 +21,22 @@ type Postgresql struct {
var ignoredColumns = map[string]bool{"datid": true, "datname": true, "stats_reset": true}
var sampleConfig = `
# specify address via a url matching:
# postgres://[pqgotest[:password]]@localhost[/dbname]?sslmode=[disable|verify-ca|verify-full]
# or a simple string:
# host=localhost user=pqotest password=... sslmode=... dbname=app_production
#
# All connection parameters are optional.
#
# Without the dbname parameter, the driver will default to a database
# with the same name as the user. This dbname is just for instantiating a
# connection with the server and doesn't restrict the databases we are trying
# to grab metrics for.
#
### specify address via a url matching:
### postgres://[pqgotest[:password]]@localhost[/dbname]?sslmode=[disable|verify-ca|verify-full]
### or a simple string:
### host=localhost user=pqotest password=... sslmode=... dbname=app_production
###
### All connection parameters are optional.
###
### Without the dbname parameter, the driver will default to a database
### with the same name as the user. This dbname is just for instantiating a
### connection with the server and doesn't restrict the databases we are trying
### to grab metrics for.
###
address = "host=localhost user=postgres sslmode=disable"
# A list of databases to pull metrics about. If not specified, metrics for all
# databases are gathered.
### A list of databases to pull metrics about. If not specified, metrics for all
### databases are gathered.
# databases = ["app_production", "testing"]
`

View File

@ -18,10 +18,8 @@ type Powerdns struct {
}
var sampleConfig = `
# An array of sockets to gather stats about.
# Specify a path to unix socket.
#
# If no servers are specified, then '/var/run/pdns.controlsocket' is used as the path.
### An array of sockets to gather stats about.
### Specify a path to unix socket.
unix_sockets = ["/var/run/pdns.controlsocket"]
`

View File

@ -30,15 +30,15 @@ func NewProcstat() *Procstat {
}
var sampleConfig = `
# Must specify one of: pid_file, exe, or pattern
# PID file to monitor process
### Must specify one of: pid_file, exe, or pattern
### PID file to monitor process
pid_file = "/var/run/nginx.pid"
# executable name (ie, pgrep <exe>)
### executable name (ie, pgrep <exe>)
# exe = "nginx"
# pattern as argument for pgrep (ie, pgrep -f <pattern>)
### pattern as argument for pgrep (ie, pgrep -f <pattern>)
# pattern = "nginx"
# Field name prefix
### Field name prefix
prefix = ""
`

View File

@ -17,7 +17,7 @@ type Prometheus struct {
}
var sampleConfig = `
# An array of urls to scrape metrics from.
### An array of urls to scrape metrics from.
urls = ["http://localhost:9100/metrics"]
`

View File

@ -18,7 +18,7 @@ type PuppetAgent struct {
}
var sampleConfig = `
# Location of puppet last run summary file
### Location of puppet last run summary file
location = "/var/lib/puppet/state/last_run_summary.yaml"
`

View File

@ -107,8 +107,8 @@ var sampleConfig = `
# username = "guest"
# password = "guest"
# A list of nodes to pull metrics about. If not specified, metrics for
# all nodes are gathered.
### A list of nodes to pull metrics about. If not specified, metrics for
### all nodes are gathered.
# nodes = ["rabbit@node1", "rabbit@node2"]
`

View File

@ -19,14 +19,14 @@ type Redis struct {
}
var sampleConfig = `
# specify servers via a url matching:
# [protocol://][:password]@address[:port]
# e.g.
# tcp://localhost:6379
# tcp://:password@192.168.99.100
#
# If no servers are specified, then localhost is used as the host.
# If no port is specified, 6379 is used
### specify servers via a url matching:
### [protocol://][:password]@address[:port]
### e.g.
### tcp://localhost:6379
### tcp://:password@192.168.99.100
###
### If no servers are specified, then localhost is used as the host.
### If no port is specified, 6379 is used
servers = ["tcp://localhost:6379"]
`

View File

@ -16,11 +16,11 @@ type RethinkDB struct {
}
var sampleConfig = `
# An array of URI to gather stats about. Specify an ip or hostname
# with optional port add password. ie rethinkdb://user:auth_key@10.10.3.30:28105,
# rethinkdb://10.10.3.33:18832, 10.0.0.1:10000, etc.
#
# If no servers are specified, then 127.0.0.1 is used as the host and 28015 as the port.
### An array of URI to gather stats about. Specify an ip or hostname
### with optional port add password. ie,
### rethinkdb://user:auth_key@10.10.3.30:28105,
### rethinkdb://10.10.3.33:18832,
### 10.0.0.1:10000, etc.
servers = ["127.0.0.1:28015"]
`

View File

@ -20,15 +20,15 @@ func (_ *Sensors) Description() string {
}
var sensorsSampleConfig = `
# By default, telegraf gathers stats from all sensors detected by the
# lm-sensors module.
#
# Only collect stats from the selected sensors. Sensors are listed as
# <chip name>:<feature name>. This information can be found by running the
# sensors command, e.g. sensors -u
#
# A * as the feature name will return all features of the chip
#
### By default, telegraf gathers stats from all sensors detected by the
### lm-sensors module.
###
### Only collect stats from the selected sensors. Sensors are listed as
### <chip name>:<feature name>. This information can be found by running the
### sensors command, e.g. sensors -u
###
### A * as the feature name will return all features of the chip
###
# sensors = ["coretemp-isa-0000:Core 0", "coretemp-isa-0001:*"]
`

View File

@ -72,11 +72,11 @@ var initNode = Node{
var NameToOid = make(map[string]string)
var sampleConfig = `
# Use 'oids.txt' file to translate oids to names
# To generate 'oids.txt' you need to run:
# snmptranslate -m all -Tz -On | sed -e 's/"//g' > /tmp/oids.txt
# Or if you have an other MIB folder with custom MIBs
# snmptranslate -M /mycustommibfolder -Tz -On -m all | sed -e 's/"//g' > oids.txt
### Use 'oids.txt' file to translate oids to names
### To generate 'oids.txt' you need to run:
### snmptranslate -m all -Tz -On | sed -e 's/"//g' > /tmp/oids.txt
### Or if you have an other MIB folder with custom MIBs
### snmptranslate -M /mycustommibfolder -Tz -On -m all | sed -e 's/"//g' > oids.txt
snmptranslate_file = "/tmp/oids.txt"
[[inputs.snmp.host]]
address = "192.168.2.2:161"

View File

@ -31,14 +31,14 @@ var queries MapQuery
var defaultServer = "Server=.;app name=telegraf;log=1;"
var sampleConfig = `
# Specify instances to monitor with a list of connection strings.
# All connection parameters are optional.
# By default, the host is localhost, listening on default port, TCP 1433.
# for Windows, the user is the currently running AD user (SSO).
# See https://github.com/denisenkom/go-mssqldb for detailed connection parameters.
### Specify instances to monitor with a list of connection strings.
### All connection parameters are optional.
### By default, the host is localhost, listening on default port, TCP 1433.
### for Windows, the user is the currently running AD user (SSO).
### See https://github.com/denisenkom/go-mssqldb for detailed connection
### parameters.
# servers = [
# "Server=192.168.1.10;Port=1433;User Id=telegraf;Password=T$l$gr@f69*;app name=telegraf;log=1;",
# "Server=192.168.1.10;Port=1433;User Id=<user>;Password=<pw>;app name=telegraf;log=1;",
# ]
`

View File

@ -11,7 +11,7 @@ import (
"sync"
"time"
"github.com/influxdata/influxdb/services/graphite"
"github.com/influxdata/telegraf/plugins/parsers/graphite"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/plugins/inputs"
@ -123,37 +123,39 @@ func (_ *Statsd) Description() string {
}
const sampleConfig = `
# Address and port to host UDP listener on
### Address and port to host UDP listener on
service_address = ":8125"
# Delete gauges every interval (default=false)
### Delete gauges every interval (default=false)
delete_gauges = false
# Delete counters every interval (default=false)
### Delete counters every interval (default=false)
delete_counters = false
# Delete sets every interval (default=false)
### Delete sets every interval (default=false)
delete_sets = false
# Delete timings & histograms every interval (default=true)
### Delete timings & histograms every interval (default=true)
delete_timings = true
# Percentiles to calculate for timing & histogram stats
### Percentiles to calculate for timing & histogram stats
percentiles = [90]
# convert measurement names, "." to "_" and "-" to "__"
### convert measurement names, "." to "_" and "-" to "__"
convert_names = true
### Statsd data translation templates, more info can be read here:
### https://github.com/influxdata/telegraf/blob/master/DATA_FORMATS_INPUT.md#graphite
# templates = [
# "cpu.* measurement*"
# ]
# Number of UDP messages allowed to queue up, once filled,
# the statsd server will start dropping packets
### Number of UDP messages allowed to queue up, once filled,
### the statsd server will start dropping packets
allowed_pending_messages = 10000
# Number of timing/histogram values to track per-measurement in the
# calculation of percentiles. Raising this limit increases the accuracy
# of percentiles but also increases the memory usage and cpu time.
### Number of timing/histogram values to track per-measurement in the
### calculation of percentiles. Raising this limit increases the accuracy
### of percentiles but also increases the memory usage and cpu time.
percentile_limit = 1000
# UDP packet size for the server to listen for. This will depend on the size
# of the packets that the client is sending, which is usually 1500 bytes.
### UDP packet size for the server to listen for. This will depend on the size
### of the packets that the client is sending, which is usually 1500 bytes.
udp_packet_size = 1500
`
@ -418,18 +420,14 @@ func (s *Statsd) parseName(bucket string) (string, string, map[string]string) {
}
}
o := graphite.Options{
Separator: "_",
Templates: s.Templates,
DefaultTags: tags,
}
var field string
name := bucketparts[0]
p, err := graphite.NewParserWithOptions(o)
p, err := graphite.NewGraphiteParser(".", s.Templates, nil)
if err == nil {
p.DefaultTags = tags
name, tags, field, _ = p.ApplyTemplate(name)
}
if s.ConvertNames {
name = strings.Replace(name, ".", "_", -1)
name = strings.Replace(name, "-", "__", -1)

View File

@ -28,11 +28,11 @@ func (_ *CPUStats) Description() string {
}
var sampleConfig = `
# Whether to report per-cpu stats or not
### Whether to report per-cpu stats or not
percpu = true
# Whether to report total system cpu stats or not
### Whether to report total system cpu stats or not
totalcpu = true
# Comment this line if you want the raw CPU time metrics
### Comment this line if you want the raw CPU time metrics
drop = ["time_*"]
`

View File

@ -21,8 +21,8 @@ func (_ *DiskStats) Description() string {
}
var diskSampleConfig = `
# By default, telegraf gather stats for all mountpoints.
# Setting mountpoints will restrict the stats to the specified mountpoints.
### By default, telegraf gather stats for all mountpoints.
### Setting mountpoints will restrict the stats to the specified mountpoints.
# mount_points = ["/"]
`

View File

@ -21,11 +21,11 @@ func (_ *NetIOStats) Description() string {
}
var netSampleConfig = `
# By default, telegraf gathers stats from any up interface (excluding loopback)
# Setting interfaces will tell it to gather these explicit interfaces,
# regardless of status.
#
# interfaces = ["eth0", ... ]
### By default, telegraf gathers stats from any up interface (excluding loopback)
### Setting interfaces will tell it to gather these explicit interfaces,
### regardless of status.
###
# interfaces = ["eth0"]
`
func (_ *NetIOStats) SampleConfig() string {

View File

@ -13,7 +13,7 @@ type NetStats struct {
}
func (_ *NetStats) Description() string {
return "Read metrics about TCP status such as established, time wait etc and UDP sockets counts."
return "Read TCP metrics such as established, time wait and sockets counts."
}
var tcpstatSampleConfig = ""

View File

@ -13,7 +13,7 @@ type Trig struct {
}
var TrigConfig = `
# Set the amplitude
### Set the amplitude
amplitude = 10.0
`
@ -42,5 +42,5 @@ func (s *Trig) Gather(acc telegraf.Accumulator) error {
}
func init() {
inputs.Add("Trig", func() telegraf.Input { return &Trig{x: 0.0} })
inputs.Add("trig", func() telegraf.Input { return &Trig{x: 0.0} })
}

View File

@ -17,9 +17,9 @@ type Twemproxy struct {
}
var sampleConfig = `
# Twemproxy stats address and port (no scheme)
### Twemproxy stats address and port (no scheme)
addr = "localhost:22222"
# Monitor pool name
### Monitor pool name
pools = ["redis_pool", "mc_pool"]
`

View File

@ -14,12 +14,12 @@ import (
)
var sampleConfig string = `
# By default this plugin returns basic CPU and Disk statistics.
# See the README file for more examples.
# Uncomment examples below or write your own as you see fit. If the system
# being polled for data does not have the Object at startup of the Telegraf
# agent, it will not be gathered.
# Settings:
### By default this plugin returns basic CPU and Disk statistics.
### See the README file for more examples.
### Uncomment examples below or write your own as you see fit. If the system
### being polled for data does not have the Object at startup of the Telegraf
### agent, it will not be gathered.
### Settings:
# PrintValid = false # Print All matching performance counters
[[inputs.win_perf_counters.object]]

View File

@ -23,16 +23,16 @@ type poolInfo struct {
}
var sampleConfig = `
# ZFS kstat path
# If not specified, then default is:
# kstatPath = "/proc/spl/kstat/zfs"
#
# By default, telegraf gather all zfs stats
# If not specified, then default is:
# kstatMetrics = ["arcstats", "zfetchstats", "vdev_cache_stats"]
#
# By default, don't gather zpool stats
# poolMetrics = false
### ZFS kstat path
### If not specified, then default is:
kstatPath = "/proc/spl/kstat/zfs"
### By default, telegraf gather all zfs stats
### If not specified, then default is:
kstatMetrics = ["arcstats", "zfetchstats", "vdev_cache_stats"]
### By default, don't gather zpool stats
poolMetrics = false
`
func (z *Zfs) SampleConfig() string {

View File

@ -20,11 +20,11 @@ type Zookeeper struct {
}
var sampleConfig = `
# An array of address to gather stats about. Specify an ip or hostname
# with port. ie localhost:2181, 10.0.0.1:2181, etc.
### An array of address to gather stats about. Specify an ip or hostname
### with port. ie localhost:2181, 10.0.0.1:2181, etc.
# If no servers are specified, then localhost is used as the host.
# If no port is specified, 2181 is used
### If no servers are specified, then localhost is used as the host.
### If no port is specified, 2181 is used
servers = [":2181"]
`

View File

@ -40,6 +40,10 @@ func TestGraphiteError(t *testing.T) {
func TestGraphiteOK(t *testing.T) {
var wg sync.WaitGroup
// Start TCP server
wg.Add(1)
go TCPServer(t, &wg)
// Init plugin
g := Graphite{
Prefix: "my.prefix",
@ -63,24 +67,15 @@ func TestGraphiteOK(t *testing.T) {
map[string]interface{}{"value": float64(3.14)},
time.Date(2010, time.November, 10, 23, 0, 0, 0, time.UTC),
)
// Prepare point list
var metrics []telegraf.Metric
metrics = append(metrics, m1)
metrics = append(metrics, m2)
metrics = append(metrics, m3)
// Start TCP server
wg.Add(1)
go TCPServer(t, &wg)
wg.Wait()
// Connect
wg.Add(1)
metrics := []telegraf.Metric{m1, m2, m3}
err1 := g.Connect()
wg.Wait()
require.NoError(t, err1)
// Send Data
err2 := g.Write(metrics)
require.NoError(t, err2)
wg.Add(1)
// Waiting TCPserver
wg.Wait()
g.Close()
@ -88,9 +83,8 @@ func TestGraphiteOK(t *testing.T) {
func TCPServer(t *testing.T, wg *sync.WaitGroup) {
tcpServer, _ := net.Listen("tcp", "127.0.0.1:2003")
wg.Done()
defer wg.Done()
conn, _ := tcpServer.Accept()
wg.Done()
reader := bufio.NewReader(conn)
tp := textproto.NewReader(reader)
data1, _ := tp.ReadLine()
@ -100,7 +94,6 @@ func TCPServer(t *testing.T, wg *sync.WaitGroup) {
data3, _ := tp.ReadLine()
assert.Equal(t, "my.prefix.192_168_0_1.my_measurement.value 3.14 1289430000", data3)
conn.Close()
wg.Done()
}
func TestGraphiteTags(t *testing.T) {

View File

@ -33,14 +33,14 @@ type InfluxDB struct {
var sampleConfig = `
### The full HTTP or UDP endpoint URL for your InfluxDB instance.
### Multiple urls can be specified but it is assumed that they are part of the same
### cluster, this means that only ONE of the urls will be written to each interval.
### Multiple urls can be specified as part of the same cluster,
### this means that only ONE of the urls will be written to each interval.
# urls = ["udp://localhost:8089"] # UDP endpoint example
urls = ["http://localhost:8086"] # required
### The target database for metrics (telegraf will create it if not exists)
database = "telegraf" # required
### Precision of writes, valid values are n, u, ms, s, m, and h
### note: using second precision greatly helps InfluxDB compression
### note: using "s" precision greatly improves InfluxDB compression
precision = "s"
### Connection timeout (for the connection with InfluxDB), formatted as a string.

View File

@ -4,6 +4,8 @@ import (
"errors"
"fmt"
"os"
"sort"
"strings"
"github.com/amir/raidman"
"github.com/influxdata/telegraf"
@ -13,6 +15,7 @@ import (
type Riemann struct {
URL string
Transport string
Separator string
client *raidman.Client
}
@ -22,6 +25,8 @@ var sampleConfig = `
url = "localhost:5555"
### transport protocol to use either tcp or udp
transport = "tcp"
### separator to use between input name and field name in Riemann service name
separator = " "
`
func (r *Riemann) Connect() error {
@ -55,7 +60,7 @@ func (r *Riemann) Write(metrics []telegraf.Metric) error {
var events []*raidman.Event
for _, p := range metrics {
evs := buildEvents(p)
evs := buildEvents(p, r.Separator)
for _, ev := range evs {
events = append(events, ev)
}
@ -70,7 +75,7 @@ func (r *Riemann) Write(metrics []telegraf.Metric) error {
return nil
}
func buildEvents(p telegraf.Metric) []*raidman.Event {
func buildEvents(p telegraf.Metric, s string) []*raidman.Event {
events := []*raidman.Event{}
for fieldName, value := range p.Fields() {
host, ok := p.Tags()["host"]
@ -85,15 +90,48 @@ func buildEvents(p telegraf.Metric) []*raidman.Event {
event := &raidman.Event{
Host: host,
Service: p.Name() + "_" + fieldName,
Metric: value,
Service: serviceName(s, p.Name(), p.Tags(), fieldName),
}
switch value.(type) {
case string:
event.State = value.(string)
default:
event.Metric = value
}
events = append(events, event)
}
return events
}
func serviceName(s string, n string, t map[string]string, f string) string {
serviceStrings := []string{}
serviceStrings = append(serviceStrings, n)
// we'll skip the 'host' tag
tagStrings := []string{}
tagNames := []string{}
for tagName := range t {
tagNames = append(tagNames, tagName)
}
sort.Strings(tagNames)
for _, tagName := range tagNames {
if tagName != "host" {
tagStrings = append(tagStrings, t[tagName])
}
}
var tagString string = strings.Join(tagStrings, s)
if tagString != "" {
serviceStrings = append(serviceStrings, tagString)
}
serviceStrings = append(serviceStrings, f)
return strings.Join(serviceStrings, s)
}
func init() {
outputs.Add("riemann", func() telegraf.Output {
return &Riemann{}

View File

@ -0,0 +1,135 @@
package graphite
import (
"fmt"
"strings"
)
const (
// DefaultSeparator is the default join character to use when joining multiple
// measurment parts in a template.
DefaultSeparator = "."
)
// Config represents the configuration for Graphite endpoints.
type Config struct {
Separator string
Templates []string
}
// Validate validates the config's templates and tags.
func (c *Config) Validate() error {
if err := c.validateTemplates(); err != nil {
return err
}
return nil
}
func (c *Config) validateTemplates() error {
// map to keep track of filters we see
filters := map[string]struct{}{}
for i, t := range c.Templates {
parts := strings.Fields(t)
// Ensure template string is non-empty
if len(parts) == 0 {
return fmt.Errorf("missing template at position: %d", i)
}
if len(parts) == 1 && parts[0] == "" {
return fmt.Errorf("missing template at position: %d", i)
}
if len(parts) > 3 {
return fmt.Errorf("invalid template format: '%s'", t)
}
template := t
filter := ""
tags := ""
if len(parts) >= 2 {
// We could have <filter> <template> or <template> <tags>. Equals is only allowed in
// tags section.
if strings.Contains(parts[1], "=") {
template = parts[0]
tags = parts[1]
} else {
filter = parts[0]
template = parts[1]
}
}
if len(parts) == 3 {
tags = parts[2]
}
// Validate the template has one and only one measurement
if err := c.validateTemplate(template); err != nil {
return err
}
// Prevent duplicate filters in the config
if _, ok := filters[filter]; ok {
return fmt.Errorf("duplicate filter '%s' found at position: %d", filter, i)
}
filters[filter] = struct{}{}
if filter != "" {
// Validate filter expression is valid
if err := c.validateFilter(filter); err != nil {
return err
}
}
if tags != "" {
// Validate tags
for _, tagStr := range strings.Split(tags, ",") {
if err := c.validateTag(tagStr); err != nil {
return err
}
}
}
}
return nil
}
func (c *Config) validateTemplate(template string) error {
hasMeasurement := false
for _, p := range strings.Split(template, ".") {
if p == "measurement" || p == "measurement*" {
hasMeasurement = true
}
}
if !hasMeasurement {
return fmt.Errorf("no measurement in template `%s`", template)
}
return nil
}
func (c *Config) validateFilter(filter string) error {
for _, p := range strings.Split(filter, ".") {
if p == "" {
return fmt.Errorf("filter contains blank section: %s", filter)
}
if strings.Contains(p, "*") && p != "*" {
return fmt.Errorf("invalid filter wildcard section: %s", filter)
}
}
return nil
}
func (c *Config) validateTag(keyValue string) error {
parts := strings.Split(keyValue, "=")
if len(parts) != 2 {
return fmt.Errorf("invalid template tags: '%s'", keyValue)
}
if parts[0] == "" || parts[1] == "" {
return fmt.Errorf("invalid template tags: %s'", keyValue)
}
return nil
}

View File

@ -0,0 +1,14 @@
package graphite
import "fmt"
// An UnsupposedValueError is returned when a parsed value is not
// supposed.
type UnsupposedValueError struct {
Field string
Value float64
}
func (err *UnsupposedValueError) Error() string {
return fmt.Sprintf(`field "%s" value: "%v" is unsupported`, err.Field, err.Value)
}

View File

@ -0,0 +1,425 @@
package graphite
import (
"bufio"
"bytes"
"fmt"
"io"
"math"
"sort"
"strconv"
"strings"
"time"
"github.com/influxdata/telegraf"
)
// Minimum and maximum supported dates for timestamps.
var (
MinDate = time.Date(1901, 12, 13, 0, 0, 0, 0, time.UTC)
MaxDate = time.Date(2038, 1, 19, 0, 0, 0, 0, time.UTC)
)
// Parser encapsulates a Graphite Parser.
type GraphiteParser struct {
Separator string
Templates []string
DefaultTags map[string]string
matcher *matcher
}
func NewGraphiteParser(
separator string,
templates []string,
defaultTags map[string]string,
) (*GraphiteParser, error) {
var err error
if separator == "" {
separator = DefaultSeparator
}
p := &GraphiteParser{
Separator: separator,
Templates: templates,
}
if defaultTags != nil {
p.DefaultTags = defaultTags
}
matcher := newMatcher()
p.matcher = matcher
defaultTemplate, _ := NewTemplate("measurement*", nil, p.Separator)
matcher.AddDefaultTemplate(defaultTemplate)
for _, pattern := range p.Templates {
template := pattern
filter := ""
// Format is [filter] <template> [tag1=value1,tag2=value2]
parts := strings.Fields(pattern)
if len(parts) < 1 {
continue
} else if len(parts) >= 2 {
if strings.Contains(parts[1], "=") {
template = parts[0]
} else {
filter = parts[0]
template = parts[1]
}
}
// Parse out the default tags specific to this template
tags := map[string]string{}
if strings.Contains(parts[len(parts)-1], "=") {
tagStrs := strings.Split(parts[len(parts)-1], ",")
for _, kv := range tagStrs {
parts := strings.Split(kv, "=")
tags[parts[0]] = parts[1]
}
}
tmpl, err1 := NewTemplate(template, tags, p.Separator)
if err1 != nil {
err = err1
break
}
matcher.Add(filter, tmpl)
}
if err != nil {
return p, fmt.Errorf("exec input parser config is error: %s ", err.Error())
} else {
return p, nil
}
}
func (p *GraphiteParser) Parse(buf []byte) ([]telegraf.Metric, error) {
// parse even if the buffer begins with a newline
buf = bytes.TrimPrefix(buf, []byte("\n"))
// add newline to end if not exists:
if len(buf) > 0 && !bytes.HasSuffix(buf, []byte("\n")) {
buf = append(buf, []byte("\n")...)
}
metrics := make([]telegraf.Metric, 0)
buffer := bytes.NewBuffer(buf)
reader := bufio.NewReader(buffer)
for {
// Read up to the next newline.
buf, err := reader.ReadBytes('\n')
if err == io.EOF {
return metrics, nil
}
if err != nil && err != io.EOF {
return metrics, err
}
// Trim the buffer, even though there should be no padding
line := strings.TrimSpace(string(buf))
if metric, err := p.ParseLine(line); err == nil {
metrics = append(metrics, metric)
}
}
}
// Parse performs Graphite parsing of a single line.
func (p *GraphiteParser) ParseLine(line string) (telegraf.Metric, error) {
// Break into 3 fields (name, value, timestamp).
fields := strings.Fields(line)
if len(fields) != 2 && len(fields) != 3 {
return nil, fmt.Errorf("received %q which doesn't have required fields", line)
}
// decode the name and tags
template := p.matcher.Match(fields[0])
measurement, tags, field, err := template.Apply(fields[0])
if err != nil {
return nil, err
}
// Could not extract measurement, use the raw value
if measurement == "" {
measurement = fields[0]
}
// Parse value.
v, err := strconv.ParseFloat(fields[1], 64)
if err != nil {
return nil, fmt.Errorf(`field "%s" value: %s`, fields[0], err)
}
if math.IsNaN(v) || math.IsInf(v, 0) {
return nil, &UnsupposedValueError{Field: fields[0], Value: v}
}
fieldValues := map[string]interface{}{}
if field != "" {
fieldValues[field] = v
} else {
fieldValues["value"] = v
}
// If no 3rd field, use now as timestamp
timestamp := time.Now().UTC()
if len(fields) == 3 {
// Parse timestamp.
unixTime, err := strconv.ParseFloat(fields[2], 64)
if err != nil {
return nil, fmt.Errorf(`field "%s" time: %s`, fields[0], err)
}
// -1 is a special value that gets converted to current UTC time
// See https://github.com/graphite-project/carbon/issues/54
if unixTime != float64(-1) {
// Check if we have fractional seconds
timestamp = time.Unix(int64(unixTime), int64((unixTime-math.Floor(unixTime))*float64(time.Second)))
if timestamp.Before(MinDate) || timestamp.After(MaxDate) {
return nil, fmt.Errorf("timestamp out of range")
}
}
}
// Set the default tags on the point if they are not already set
for k, v := range p.DefaultTags {
if _, ok := tags[k]; !ok {
tags[k] = v
}
}
return telegraf.NewMetric(measurement, tags, fieldValues, timestamp)
}
// ApplyTemplate extracts the template fields from the given line and
// returns the measurement name and tags.
func (p *GraphiteParser) ApplyTemplate(line string) (string, map[string]string, string, error) {
// Break line into fields (name, value, timestamp), only name is used
fields := strings.Fields(line)
if len(fields) == 0 {
return "", make(map[string]string), "", nil
}
// decode the name and tags
template := p.matcher.Match(fields[0])
name, tags, field, err := template.Apply(fields[0])
// Set the default tags on the point if they are not already set
for k, v := range p.DefaultTags {
if _, ok := tags[k]; !ok {
tags[k] = v
}
}
return name, tags, field, err
}
// template represents a pattern and tags to map a graphite metric string to a influxdb Point
type template struct {
tags []string
defaultTags map[string]string
greedyMeasurement bool
separator string
}
// NewTemplate returns a new template ensuring it has a measurement
// specified.
func NewTemplate(pattern string, defaultTags map[string]string, separator string) (*template, error) {
tags := strings.Split(pattern, ".")
hasMeasurement := false
template := &template{tags: tags, defaultTags: defaultTags, separator: separator}
for _, tag := range tags {
if strings.HasPrefix(tag, "measurement") {
hasMeasurement = true
}
if tag == "measurement*" {
template.greedyMeasurement = true
}
}
if !hasMeasurement {
return nil, fmt.Errorf("no measurement specified for template. %q", pattern)
}
return template, nil
}
// Apply extracts the template fields from the given line and returns the measurement
// name and tags
func (t *template) Apply(line string) (string, map[string]string, string, error) {
fields := strings.Split(line, ".")
var (
measurement []string
tags = make(map[string]string)
field string
)
// Set any default tags
for k, v := range t.defaultTags {
tags[k] = v
}
for i, tag := range t.tags {
if i >= len(fields) {
continue
}
if tag == "measurement" {
measurement = append(measurement, fields[i])
} else if tag == "field" {
if len(field) != 0 {
return "", nil, "", fmt.Errorf("'field' can only be used once in each template: %q", line)
}
field = fields[i]
} else if tag == "measurement*" {
measurement = append(measurement, fields[i:]...)
break
} else if tag != "" {
tags[tag] = fields[i]
}
}
return strings.Join(measurement, t.separator), tags, field, nil
}
// matcher determines which template should be applied to a given metric
// based on a filter tree.
type matcher struct {
root *node
defaultTemplate *template
}
func newMatcher() *matcher {
return &matcher{
root: &node{},
}
}
// Add inserts the template in the filter tree based the given filter
func (m *matcher) Add(filter string, template *template) {
if filter == "" {
m.AddDefaultTemplate(template)
return
}
m.root.Insert(filter, template)
}
func (m *matcher) AddDefaultTemplate(template *template) {
m.defaultTemplate = template
}
// Match returns the template that matches the given graphite line
func (m *matcher) Match(line string) *template {
tmpl := m.root.Search(line)
if tmpl != nil {
return tmpl
}
return m.defaultTemplate
}
// node is an item in a sorted k-ary tree. Each child is sorted by its value.
// The special value of "*", is always last.
type node struct {
value string
children nodes
template *template
}
func (n *node) insert(values []string, template *template) {
// Add the end, set the template
if len(values) == 0 {
n.template = template
return
}
// See if the the current element already exists in the tree. If so, insert the
// into that sub-tree
for _, v := range n.children {
if v.value == values[0] {
v.insert(values[1:], template)
return
}
}
// New element, add it to the tree and sort the children
newNode := &node{value: values[0]}
n.children = append(n.children, newNode)
sort.Sort(&n.children)
// Now insert the rest of the tree into the new element
newNode.insert(values[1:], template)
}
// Insert inserts the given string template into the tree. The filter string is separated
// on "." and each part is used as the path in the tree.
func (n *node) Insert(filter string, template *template) {
n.insert(strings.Split(filter, "."), template)
}
func (n *node) search(lineParts []string) *template {
// Nothing to search
if len(lineParts) == 0 || len(n.children) == 0 {
return n.template
}
// If last element is a wildcard, don't include in this search since it's sorted
// to the end but lexicographically it would not always be and sort.Search assumes
// the slice is sorted.
length := len(n.children)
if n.children[length-1].value == "*" {
length--
}
// Find the index of child with an exact match
i := sort.Search(length, func(i int) bool {
return n.children[i].value >= lineParts[0]
})
// Found an exact match, so search that child sub-tree
if i < len(n.children) && n.children[i].value == lineParts[0] {
return n.children[i].search(lineParts[1:])
}
// Not an exact match, see if we have a wildcard child to search
if n.children[len(n.children)-1].value == "*" {
return n.children[len(n.children)-1].search(lineParts[1:])
}
return n.template
}
func (n *node) Search(line string) *template {
return n.search(strings.Split(line, "."))
}
type nodes []*node
// Less returns a boolean indicating whether the filter at position j
// is less than the filter at position k. Filters are order by string
// comparison of each component parts. A wildcard value "*" is never
// less than a non-wildcard value.
//
// For example, the filters:
// "*.*"
// "servers.*"
// "servers.localhost"
// "*.localhost"
//
// Would be sorted as:
// "servers.localhost"
// "servers.*"
// "*.localhost"
// "*.*"
func (n *nodes) Less(j, k int) bool {
if (*n)[j].value == "*" && (*n)[k].value != "*" {
return false
}
if (*n)[j].value != "*" && (*n)[k].value == "*" {
return true
}
return (*n)[j].value < (*n)[k].value
}
func (n *nodes) Swap(i, j int) { (*n)[i], (*n)[j] = (*n)[j], (*n)[i] }
func (n *nodes) Len() int { return len(*n) }

View File

@ -0,0 +1,595 @@
package graphite
import (
"reflect"
"strconv"
"testing"
"time"
"github.com/influxdata/telegraf"
"github.com/stretchr/testify/assert"
)
func BenchmarkParse(b *testing.B) {
p, err := NewGraphiteParser("_", []string{
"*.* .wrong.measurement*",
"servers.* .host.measurement*",
"servers.localhost .host.measurement*",
"*.localhost .host.measurement*",
"*.*.cpu .host.measurement*",
"a.b.c .host.measurement*",
"influxd.*.foo .host.measurement*",
"prod.*.mem .host.measurement*",
}, nil)
if err != nil {
b.Fatalf("unexpected error creating parser, got %v", err)
}
for i := 0; i < b.N; i++ {
p.Parse([]byte("servers.localhost.cpu.load 11 1435077219"))
}
}
func TestTemplateApply(t *testing.T) {
var tests = []struct {
test string
input string
template string
measurement string
tags map[string]string
err string
}{
{
test: "metric only",
input: "cpu",
template: "measurement",
measurement: "cpu",
},
{
test: "metric with single series",
input: "cpu.server01",
template: "measurement.hostname",
measurement: "cpu",
tags: map[string]string{"hostname": "server01"},
},
{
test: "metric with multiple series",
input: "cpu.us-west.server01",
template: "measurement.region.hostname",
measurement: "cpu",
tags: map[string]string{"hostname": "server01", "region": "us-west"},
},
{
test: "no metric",
tags: make(map[string]string),
err: `no measurement specified for template. ""`,
},
{
test: "ignore unnamed",
input: "foo.cpu",
template: "measurement",
measurement: "foo",
tags: make(map[string]string),
},
{
test: "name shorter than template",
input: "foo",
template: "measurement.A.B.C",
measurement: "foo",
tags: make(map[string]string),
},
{
test: "wildcard measurement at end",
input: "prod.us-west.server01.cpu.load",
template: "env.zone.host.measurement*",
measurement: "cpu.load",
tags: map[string]string{"env": "prod", "zone": "us-west", "host": "server01"},
},
{
test: "skip fields",
input: "ignore.us-west.ignore-this-too.cpu.load",
template: ".zone..measurement*",
measurement: "cpu.load",
tags: map[string]string{"zone": "us-west"},
},
}
for _, test := range tests {
tmpl, err := NewTemplate(test.template, nil, DefaultSeparator)
if errstr(err) != test.err {
t.Fatalf("err does not match. expected %v, got %v", test.err, err)
}
if err != nil {
// If we erred out,it was intended and the following tests won't work
continue
}
measurement, tags, _, _ := tmpl.Apply(test.input)
if measurement != test.measurement {
t.Fatalf("name parse failer. expected %v, got %v", test.measurement, measurement)
}
if len(tags) != len(test.tags) {
t.Fatalf("unexpected number of tags. expected %v, got %v", test.tags, tags)
}
for k, v := range test.tags {
if tags[k] != v {
t.Fatalf("unexpected tag value for tags[%s]. expected %q, got %q", k, v, tags[k])
}
}
}
}
func TestParseMissingMeasurement(t *testing.T) {
_, err := NewGraphiteParser("", []string{"a.b.c"}, nil)
if err == nil {
t.Fatalf("expected error creating parser, got nil")
}
}
func TestParse(t *testing.T) {
testTime := time.Now().Round(time.Second)
epochTime := testTime.Unix()
strTime := strconv.FormatInt(epochTime, 10)
var tests = []struct {
test string
input string
measurement string
tags map[string]string
value float64
time time.Time
template string
err string
}{
{
test: "normal case",
input: `cpu.foo.bar 50 ` + strTime,
template: "measurement.foo.bar",
measurement: "cpu",
tags: map[string]string{
"foo": "foo",
"bar": "bar",
},
value: 50,
time: testTime,
},
{
test: "metric only with float value",
input: `cpu 50.554 ` + strTime,
measurement: "cpu",
template: "measurement",
value: 50.554,
time: testTime,
},
{
test: "missing metric",
input: `1419972457825`,
template: "measurement",
err: `received "1419972457825" which doesn't have required fields`,
},
{
test: "should error parsing invalid float",
input: `cpu 50.554z 1419972457825`,
template: "measurement",
err: `field "cpu" value: strconv.ParseFloat: parsing "50.554z": invalid syntax`,
},
{
test: "should error parsing invalid int",
input: `cpu 50z 1419972457825`,
template: "measurement",
err: `field "cpu" value: strconv.ParseFloat: parsing "50z": invalid syntax`,
},
{
test: "should error parsing invalid time",
input: `cpu 50.554 14199724z57825`,
template: "measurement",
err: `field "cpu" time: strconv.ParseFloat: parsing "14199724z57825": invalid syntax`,
},
}
for _, test := range tests {
p, err := NewGraphiteParser("", []string{test.template}, nil)
if err != nil {
t.Fatalf("unexpected error creating graphite parser: %v", err)
}
metric, err := p.ParseLine(test.input)
if errstr(err) != test.err {
t.Fatalf("err does not match. expected %v, got %v", test.err, err)
}
if err != nil {
// If we erred out,it was intended and the following tests won't work
continue
}
if metric.Name() != test.measurement {
t.Fatalf("name parse failer. expected %v, got %v",
test.measurement, metric.Name())
}
if len(metric.Tags()) != len(test.tags) {
t.Fatalf("tags len mismatch. expected %d, got %d",
len(test.tags), len(metric.Tags()))
}
f := metric.Fields()["value"].(float64)
if metric.Fields()["value"] != f {
t.Fatalf("floatValue value mismatch. expected %v, got %v",
test.value, f)
}
if metric.Time().UnixNano()/1000000 != test.time.UnixNano()/1000000 {
t.Fatalf("time value mismatch. expected %v, got %v",
test.time.UnixNano(), metric.Time().UnixNano())
}
}
}
func TestParseNaN(t *testing.T) {
p, err := NewGraphiteParser("", []string{"measurement*"}, nil)
assert.NoError(t, err)
_, err = p.ParseLine("servers.localhost.cpu_load NaN 1435077219")
assert.Error(t, err)
if _, ok := err.(*UnsupposedValueError); !ok {
t.Fatalf("expected *ErrUnsupportedValue, got %v", reflect.TypeOf(err))
}
}
func TestFilterMatchDefault(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.localhost .host.measurement*"}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("miss.servers.localhost.cpu_load",
map[string]string{},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("miss.servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestFilterMatchMultipleMeasurement(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.localhost .host.measurement.measurement*"}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu.cpu_load.10",
map[string]string{"host": "localhost"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu.cpu_load.10 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestFilterMatchMultipleMeasurementSeparator(t *testing.T) {
p, err := NewGraphiteParser("_",
[]string{"servers.localhost .host.measurement.measurement*"},
nil,
)
assert.NoError(t, err)
exp, err := telegraf.NewMetric("cpu_cpu_load_10",
map[string]string{"host": "localhost"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu.cpu_load.10 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestFilterMatchSingle(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.localhost .host.measurement*"}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
m, err := p.ParseLine("servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestParseNoMatch(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.*.cpu .host.measurement.cpu.measurement"}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("servers.localhost.memory.VmallocChunk",
map[string]string{},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.memory.VmallocChunk 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestFilterMatchWildcard(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.* .host.measurement*"}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestFilterMatchExactBeforeWildcard(t *testing.T) {
p, err := NewGraphiteParser("", []string{
"servers.* .wrong.measurement*",
"servers.localhost .host.measurement*"}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestFilterMatchMostLongestFilter(t *testing.T) {
p, err := NewGraphiteParser("", []string{
"*.* .wrong.measurement*",
"servers.* .wrong.measurement*",
"servers.localhost .wrong.measurement*",
"servers.localhost.cpu .host.resource.measurement*", // should match this
"*.localhost .wrong.measurement*",
}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost", "resource": "cpu"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestFilterMatchMultipleWildcards(t *testing.T) {
p, err := NewGraphiteParser("", []string{
"*.* .wrong.measurement*",
"servers.* .host.measurement*", // should match this
"servers.localhost .wrong.measurement*",
"*.localhost .wrong.measurement*",
}, nil)
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "server01"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.server01.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestParseDefaultTags(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.localhost .host.measurement*"}, map[string]string{
"region": "us-east",
"zone": "1c",
"host": "should not set",
})
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost", "region": "us-east", "zone": "1c"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestParseDefaultTemplateTags(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.localhost .host.measurement* zone=1c"}, map[string]string{
"region": "us-east",
"host": "should not set",
})
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost", "region": "us-east", "zone": "1c"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestParseDefaultTemplateTagsOverridGlobal(t *testing.T) {
p, err := NewGraphiteParser("", []string{"servers.localhost .host.measurement* zone=1c,region=us-east"}, map[string]string{
"region": "shot not be set",
"host": "should not set",
})
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost", "region": "us-east", "zone": "1c"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
func TestParseTemplateWhitespace(t *testing.T) {
p, err := NewGraphiteParser("",
[]string{"servers.localhost .host.measurement* zone=1c"},
map[string]string{
"region": "us-east",
"host": "should not set",
})
if err != nil {
t.Fatalf("unexpected error creating parser, got %v", err)
}
exp, err := telegraf.NewMetric("cpu_load",
map[string]string{"host": "localhost", "region": "us-east", "zone": "1c"},
map[string]interface{}{"value": float64(11)},
time.Unix(1435077219, 0))
assert.NoError(t, err)
m, err := p.ParseLine("servers.localhost.cpu_load 11 1435077219")
assert.NoError(t, err)
assert.Equal(t, exp.String(), m.String())
}
// Test basic functionality of ApplyTemplate
func TestApplyTemplate(t *testing.T) {
p, err := NewGraphiteParser("_",
[]string{"current.* measurement.measurement"},
nil)
assert.NoError(t, err)
measurement, _, _, _ := p.ApplyTemplate("current.users")
assert.Equal(t, "current_users", measurement)
}
// Test basic functionality of ApplyTemplate
func TestApplyTemplateNoMatch(t *testing.T) {
p, err := NewGraphiteParser(".",
[]string{"foo.bar measurement.measurement"},
nil)
assert.NoError(t, err)
measurement, _, _, _ := p.ApplyTemplate("current.users")
assert.Equal(t, "current.users", measurement)
}
// Test that most specific template is chosen
func TestApplyTemplateSpecific(t *testing.T) {
p, err := NewGraphiteParser("_",
[]string{
"current.* measurement.measurement",
"current.*.* measurement.measurement.service",
}, nil)
assert.NoError(t, err)
measurement, tags, _, _ := p.ApplyTemplate("current.users.facebook")
assert.Equal(t, "current_users", measurement)
service, ok := tags["service"]
if !ok {
t.Error("Expected for template to apply a 'service' tag, but not found")
}
if service != "facebook" {
t.Errorf("Expected service='facebook' tag, got service='%s'", service)
}
}
func TestApplyTemplateTags(t *testing.T) {
p, err := NewGraphiteParser("_",
[]string{"current.* measurement.measurement region=us-west"}, nil)
assert.NoError(t, err)
measurement, tags, _, _ := p.ApplyTemplate("current.users")
assert.Equal(t, "current_users", measurement)
region, ok := tags["region"]
if !ok {
t.Error("Expected for template to apply a 'region' tag, but not found")
}
if region != "us-west" {
t.Errorf("Expected region='us-west' tag, got region='%s'", region)
}
}
func TestApplyTemplateField(t *testing.T) {
p, err := NewGraphiteParser("_",
[]string{"current.* measurement.measurement.field"}, nil)
assert.NoError(t, err)
measurement, _, field, err := p.ApplyTemplate("current.users.logged_in")
assert.Equal(t, "current_users", measurement)
if field != "logged_in" {
t.Errorf("Parser.ApplyTemplate unexpected result. got %s, exp %s",
field, "logged_in")
}
}
func TestApplyTemplateFieldError(t *testing.T) {
p, err := NewGraphiteParser("_",
[]string{"current.* measurement.field.field"}, nil)
assert.NoError(t, err)
_, _, _, err = p.ApplyTemplate("current.users.logged_in")
if err == nil {
t.Errorf("Parser.ApplyTemplate unexpected result. got %s, exp %s", err,
"'field' can only be used once in each template: current.users.logged_in")
}
}
// Test Helpers
func errstr(err error) string {
if err != nil {
return err.Error()
}
return ""
}

View File

@ -0,0 +1,57 @@
package influx
import (
"bytes"
"fmt"
"github.com/influxdata/telegraf"
"github.com/influxdata/influxdb/models"
)
// InfluxParser is an object for Parsing incoming metrics.
type InfluxParser struct {
// DefaultTags will be added to every parsed metric
DefaultTags map[string]string
}
// ParseMetrics returns a slice of Metrics from a text representation of a
// metric (in line-protocol format)
// with each metric separated by newlines. If any metrics fail to parse,
// a non-nil error will be returned in addition to the metrics that parsed
// successfully.
func (p *InfluxParser) Parse(buf []byte) ([]telegraf.Metric, error) {
// parse even if the buffer begins with a newline
buf = bytes.TrimPrefix(buf, []byte("\n"))
points, err := models.ParsePoints(buf)
metrics := make([]telegraf.Metric, len(points))
for i, point := range points {
tags := point.Tags()
for k, v := range p.DefaultTags {
// Only set tags not in parsed metric
if _, ok := tags[k]; !ok {
tags[k] = v
}
}
// Ignore error here because it's impossible that a model.Point
// wouldn't parse into client.Point properly
metrics[i], _ = telegraf.NewMetric(point.Name(), tags,
point.Fields(), point.Time())
}
return metrics, err
}
func (p *InfluxParser) ParseLine(line string) (telegraf.Metric, error) {
metrics, err := p.Parse([]byte(line + "\n"))
if err != nil {
return nil, err
}
if len(metrics) < 1 {
return nil, fmt.Errorf(
"Can not parse the line: %s, for data format: influx ", line)
}
return metrics[0], nil
}

View File

@ -0,0 +1,194 @@
package influx
import (
"testing"
"time"
"github.com/stretchr/testify/assert"
)
var exptime = time.Date(2009, time.November, 10, 23, 0, 0, 0, time.UTC)
const (
validInflux = "cpu_load_short,cpu=cpu0 value=10 1257894000000000000"
validInfluxNewline = "\ncpu_load_short,cpu=cpu0 value=10 1257894000000000000\n"
invalidInflux = "I don't think this is line protocol"
invalidInflux2 = "{\"a\": 5, \"b\": {\"c\": 6}}"
)
const influxMulti = `
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
`
const influxMultiSomeInvalid = `
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,cpu=cpu3, host=foo,datacenter=us-east usage_idle=99,usage_busy=1
cpu,cpu=cpu4 , usage_idle=99,usage_busy=1
cpu,host=foo,datacenter=us-east usage_idle=99,usage_busy=1
`
func TestParseValidInflux(t *testing.T) {
parser := InfluxParser{}
metrics, err := parser.Parse([]byte(validInflux))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "cpu_load_short", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"value": float64(10),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{
"cpu": "cpu0",
}, metrics[0].Tags())
assert.Equal(t, exptime, metrics[0].Time())
metrics, err = parser.Parse([]byte(validInfluxNewline))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "cpu_load_short", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"value": float64(10),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{
"cpu": "cpu0",
}, metrics[0].Tags())
assert.Equal(t, exptime, metrics[0].Time())
}
func TestParseLineValidInflux(t *testing.T) {
parser := InfluxParser{}
metric, err := parser.ParseLine(validInflux)
assert.NoError(t, err)
assert.Equal(t, "cpu_load_short", metric.Name())
assert.Equal(t, map[string]interface{}{
"value": float64(10),
}, metric.Fields())
assert.Equal(t, map[string]string{
"cpu": "cpu0",
}, metric.Tags())
assert.Equal(t, exptime, metric.Time())
metric, err = parser.ParseLine(validInfluxNewline)
assert.NoError(t, err)
assert.Equal(t, "cpu_load_short", metric.Name())
assert.Equal(t, map[string]interface{}{
"value": float64(10),
}, metric.Fields())
assert.Equal(t, map[string]string{
"cpu": "cpu0",
}, metric.Tags())
assert.Equal(t, exptime, metric.Time())
}
func TestParseMultipleValid(t *testing.T) {
parser := InfluxParser{}
metrics, err := parser.Parse([]byte(influxMulti))
assert.NoError(t, err)
assert.Len(t, metrics, 7)
for _, metric := range metrics {
assert.Equal(t, "cpu", metric.Name())
assert.Equal(t, map[string]string{
"datacenter": "us-east",
"host": "foo",
}, metrics[0].Tags())
assert.Equal(t, map[string]interface{}{
"usage_idle": float64(99),
"usage_busy": float64(1),
}, metrics[0].Fields())
}
}
func TestParseSomeValid(t *testing.T) {
parser := InfluxParser{}
metrics, err := parser.Parse([]byte(influxMultiSomeInvalid))
assert.Error(t, err)
assert.Len(t, metrics, 4)
for _, metric := range metrics {
assert.Equal(t, "cpu", metric.Name())
assert.Equal(t, map[string]string{
"datacenter": "us-east",
"host": "foo",
}, metrics[0].Tags())
assert.Equal(t, map[string]interface{}{
"usage_idle": float64(99),
"usage_busy": float64(1),
}, metrics[0].Fields())
}
}
// Test that default tags are applied.
func TestParseDefaultTags(t *testing.T) {
parser := InfluxParser{
DefaultTags: map[string]string{
"tag": "default",
},
}
metrics, err := parser.Parse([]byte(influxMultiSomeInvalid))
assert.Error(t, err)
assert.Len(t, metrics, 4)
for _, metric := range metrics {
assert.Equal(t, "cpu", metric.Name())
assert.Equal(t, map[string]string{
"datacenter": "us-east",
"host": "foo",
"tag": "default",
}, metrics[0].Tags())
assert.Equal(t, map[string]interface{}{
"usage_idle": float64(99),
"usage_busy": float64(1),
}, metrics[0].Fields())
}
}
// Verify that metric tags will override default tags
func TestParseDefaultTagsOverride(t *testing.T) {
parser := InfluxParser{
DefaultTags: map[string]string{
"host": "default",
},
}
metrics, err := parser.Parse([]byte(influxMultiSomeInvalid))
assert.Error(t, err)
assert.Len(t, metrics, 4)
for _, metric := range metrics {
assert.Equal(t, "cpu", metric.Name())
assert.Equal(t, map[string]string{
"datacenter": "us-east",
"host": "foo",
}, metrics[0].Tags())
assert.Equal(t, map[string]interface{}{
"usage_idle": float64(99),
"usage_busy": float64(1),
}, metrics[0].Fields())
}
}
func TestParseInvalidInflux(t *testing.T) {
parser := InfluxParser{}
_, err := parser.Parse([]byte(invalidInflux))
assert.Error(t, err)
_, err = parser.Parse([]byte(invalidInflux2))
assert.Error(t, err)
_, err = parser.ParseLine(invalidInflux)
assert.Error(t, err)
_, err = parser.ParseLine(invalidInflux2)
assert.Error(t, err)
}

View File

@ -0,0 +1,109 @@
package json
import (
"encoding/json"
"fmt"
"strconv"
"strings"
"time"
"github.com/influxdata/telegraf"
)
type JSONParser struct {
MetricName string
TagKeys []string
DefaultTags map[string]string
}
func (p *JSONParser) Parse(buf []byte) ([]telegraf.Metric, error) {
metrics := make([]telegraf.Metric, 0)
var jsonOut map[string]interface{}
err := json.Unmarshal(buf, &jsonOut)
if err != nil {
err = fmt.Errorf("unable to parse out as JSON, %s", err)
return nil, err
}
tags := make(map[string]string)
for k, v := range p.DefaultTags {
tags[k] = v
}
for _, tag := range p.TagKeys {
switch v := jsonOut[tag].(type) {
case string:
tags[tag] = v
}
delete(jsonOut, tag)
}
f := JSONFlattener{}
err = f.FlattenJSON("", jsonOut)
if err != nil {
return nil, err
}
metric, err := telegraf.NewMetric(p.MetricName, tags, f.Fields, time.Now().UTC())
if err != nil {
return nil, err
}
return append(metrics, metric), nil
}
func (p *JSONParser) ParseLine(line string) (telegraf.Metric, error) {
metrics, err := p.Parse([]byte(line + "\n"))
if err != nil {
return nil, err
}
if len(metrics) < 1 {
return nil, fmt.Errorf("Can not parse the line: %s, for data format: influx ", line)
}
return metrics[0], nil
}
type JSONFlattener struct {
Fields map[string]interface{}
}
// FlattenJSON flattens nested maps/interfaces into a fields map
func (f *JSONFlattener) FlattenJSON(
fieldname string,
v interface{},
) error {
if f.Fields == nil {
f.Fields = make(map[string]interface{})
}
fieldname = strings.Trim(fieldname, "_")
switch t := v.(type) {
case map[string]interface{}:
for k, v := range t {
err := f.FlattenJSON(fieldname+"_"+k+"_", v)
if err != nil {
return err
}
}
case []interface{}:
for i, v := range t {
k := strconv.Itoa(i)
err := f.FlattenJSON(fieldname+"_"+k+"_", v)
if err != nil {
return nil
}
}
case float64:
f.Fields[fieldname] = t
case bool, string, nil:
// ignored types
return nil
default:
return fmt.Errorf("JSON Flattener: got unexpected type %T with value %v (%s)",
t, t, fieldname)
}
return nil
}

View File

@ -0,0 +1,284 @@
package json
import (
"testing"
"github.com/stretchr/testify/assert"
)
const (
validJSON = "{\"a\": 5, \"b\": {\"c\": 6}}"
validJSONNewline = "\n{\"d\": 7, \"b\": {\"d\": 8}}\n"
invalidJSON = "I don't think this is JSON"
invalidJSON2 = "{\"a\": 5, \"b\": \"c\": 6}}"
)
const validJSONTags = `
{
"a": 5,
"b": {
"c": 6
},
"mytag": "foobar",
"othertag": "baz"
}
`
func TestParseValidJSON(t *testing.T) {
parser := JSONParser{
MetricName: "json_test",
}
// Most basic vanilla test
metrics, err := parser.Parse([]byte(validJSON))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{}, metrics[0].Tags())
// Test that newlines are fine
metrics, err = parser.Parse([]byte(validJSONNewline))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"d": float64(7),
"b_d": float64(8),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{}, metrics[0].Tags())
// Test that strings without TagKeys defined are ignored
metrics, err = parser.Parse([]byte(validJSONTags))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{}, metrics[0].Tags())
}
func TestParseLineValidJSON(t *testing.T) {
parser := JSONParser{
MetricName: "json_test",
}
// Most basic vanilla test
metric, err := parser.ParseLine(validJSON)
assert.NoError(t, err)
assert.Equal(t, "json_test", metric.Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metric.Fields())
assert.Equal(t, map[string]string{}, metric.Tags())
// Test that newlines are fine
metric, err = parser.ParseLine(validJSONNewline)
assert.NoError(t, err)
assert.Equal(t, "json_test", metric.Name())
assert.Equal(t, map[string]interface{}{
"d": float64(7),
"b_d": float64(8),
}, metric.Fields())
assert.Equal(t, map[string]string{}, metric.Tags())
// Test that strings without TagKeys defined are ignored
metric, err = parser.ParseLine(validJSONTags)
assert.NoError(t, err)
assert.Equal(t, "json_test", metric.Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metric.Fields())
assert.Equal(t, map[string]string{}, metric.Tags())
}
func TestParseInvalidJSON(t *testing.T) {
parser := JSONParser{
MetricName: "json_test",
}
_, err := parser.Parse([]byte(invalidJSON))
assert.Error(t, err)
_, err = parser.Parse([]byte(invalidJSON2))
assert.Error(t, err)
_, err = parser.ParseLine(invalidJSON)
assert.Error(t, err)
}
func TestParseWithTagKeys(t *testing.T) {
// Test that strings not matching tag keys are ignored
parser := JSONParser{
MetricName: "json_test",
TagKeys: []string{"wrongtagkey"},
}
metrics, err := parser.Parse([]byte(validJSONTags))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{}, metrics[0].Tags())
// Test that single tag key is found and applied
parser = JSONParser{
MetricName: "json_test",
TagKeys: []string{"mytag"},
}
metrics, err = parser.Parse([]byte(validJSONTags))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{
"mytag": "foobar",
}, metrics[0].Tags())
// Test that both tag keys are found and applied
parser = JSONParser{
MetricName: "json_test",
TagKeys: []string{"mytag", "othertag"},
}
metrics, err = parser.Parse([]byte(validJSONTags))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{
"mytag": "foobar",
"othertag": "baz",
}, metrics[0].Tags())
}
func TestParseLineWithTagKeys(t *testing.T) {
// Test that strings not matching tag keys are ignored
parser := JSONParser{
MetricName: "json_test",
TagKeys: []string{"wrongtagkey"},
}
metric, err := parser.ParseLine(validJSONTags)
assert.NoError(t, err)
assert.Equal(t, "json_test", metric.Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metric.Fields())
assert.Equal(t, map[string]string{}, metric.Tags())
// Test that single tag key is found and applied
parser = JSONParser{
MetricName: "json_test",
TagKeys: []string{"mytag"},
}
metric, err = parser.ParseLine(validJSONTags)
assert.NoError(t, err)
assert.Equal(t, "json_test", metric.Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metric.Fields())
assert.Equal(t, map[string]string{
"mytag": "foobar",
}, metric.Tags())
// Test that both tag keys are found and applied
parser = JSONParser{
MetricName: "json_test",
TagKeys: []string{"mytag", "othertag"},
}
metric, err = parser.ParseLine(validJSONTags)
assert.NoError(t, err)
assert.Equal(t, "json_test", metric.Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metric.Fields())
assert.Equal(t, map[string]string{
"mytag": "foobar",
"othertag": "baz",
}, metric.Tags())
}
func TestParseValidJSONDefaultTags(t *testing.T) {
parser := JSONParser{
MetricName: "json_test",
TagKeys: []string{"mytag"},
DefaultTags: map[string]string{
"t4g": "default",
},
}
// Most basic vanilla test
metrics, err := parser.Parse([]byte(validJSON))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{"t4g": "default"}, metrics[0].Tags())
// Test that tagkeys and default tags are applied
metrics, err = parser.Parse([]byte(validJSONTags))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{
"t4g": "default",
"mytag": "foobar",
}, metrics[0].Tags())
}
// Test that default tags are overridden by tag keys
func TestParseValidJSONDefaultTagsOverride(t *testing.T) {
parser := JSONParser{
MetricName: "json_test",
TagKeys: []string{"mytag"},
DefaultTags: map[string]string{
"mytag": "default",
},
}
// Most basic vanilla test
metrics, err := parser.Parse([]byte(validJSON))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{"mytag": "default"}, metrics[0].Tags())
// Test that tagkeys override default tags
metrics, err = parser.Parse([]byte(validJSONTags))
assert.NoError(t, err)
assert.Len(t, metrics, 1)
assert.Equal(t, "json_test", metrics[0].Name())
assert.Equal(t, map[string]interface{}{
"a": float64(5),
"b_c": float64(6),
}, metrics[0].Fields())
assert.Equal(t, map[string]string{
"mytag": "foobar",
}, metrics[0].Tags())
}

View File

@ -0,0 +1,95 @@
package parsers
import (
"fmt"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/plugins/parsers/graphite"
"github.com/influxdata/telegraf/plugins/parsers/influx"
"github.com/influxdata/telegraf/plugins/parsers/json"
)
// ParserInput is an interface for input plugins that are able to parse
// arbitrary data formats.
type ParserInput interface {
// SetParser sets the parser function for the interface
SetParser(parser Parser)
}
// Parser is an interface defining functions that a parser plugin must satisfy.
type Parser interface {
// Parse takes a byte buffer separated by newlines
// ie, `cpu.usage.idle 90\ncpu.usage.busy 10`
// and parses it into telegraf metrics
Parse(buf []byte) ([]telegraf.Metric, error)
// ParseLine takes a single string metric
// ie, "cpu.usage.idle 90"
// and parses it into a telegraf metric.
ParseLine(line string) (telegraf.Metric, error)
}
// Config is a struct that covers the data types needed for all parser types,
// and can be used to instantiate _any_ of the parsers.
type Config struct {
// Dataformat can be one of: json, influx, graphite
DataFormat string
// Separator only applied to Graphite data.
Separator string
// Templates only apply to Graphite data.
Templates []string
// TagKeys only apply to JSON data
TagKeys []string
// MetricName only applies to JSON data. This will be the name of the measurement.
MetricName string
// DefaultTags are the default tags that will be added to all parsed metrics.
DefaultTags map[string]string
}
// NewParser returns a Parser interface based on the given config.
func NewParser(config *Config) (Parser, error) {
var err error
var parser Parser
switch config.DataFormat {
case "json":
parser, err = NewJSONParser(config.MetricName,
config.TagKeys, config.DefaultTags)
case "influx":
parser, err = NewInfluxParser()
case "graphite":
parser, err = NewGraphiteParser(config.Separator,
config.Templates, config.DefaultTags)
default:
err = fmt.Errorf("Invalid data format: %s", config.DataFormat)
}
return parser, err
}
func NewJSONParser(
metricName string,
tagKeys []string,
defaultTags map[string]string,
) (Parser, error) {
parser := &json.JSONParser{
MetricName: metricName,
TagKeys: tagKeys,
DefaultTags: defaultTags,
}
return parser, nil
}
func NewInfluxParser() (Parser, error) {
return &influx.InfluxParser{}, nil
}
func NewGraphiteParser(
separator string,
templates []string,
defaultTags map[string]string,
) (Parser, error) {
return graphite.NewGraphiteParser(separator, templates, defaultTags)
}

View File

@ -68,4 +68,6 @@ tmpdir=$(mktemp -d)
exit_if_fail ./telegraf -config $tmpdir/config.toml \
-test -input-filter cpu:mem
mv ./telegraf $CIRCLE_ARTIFACTS
exit $rc