dom
/
telegraf
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Fix uptime_ns calculation when container has been restarted (#6649)

This commit is contained in:
Daniel Nelson 2019-11-12 11:58:13 -08:00 committed by GitHub
parent d858d82a85
commit ce3ae58ad9
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 143 additions and 74 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
plugins/inputs/docker/docker.go
View File

@ -546,17 +546,22 @@ func (d *Docker) gatherContainerInspect(
started, err := time.Parse(time.RFC3339, info.State.StartedAt) started, err := time.Parse(time.RFC3339, info.State.StartedAt)
if err == nil && !started.IsZero() { if err == nil && !started.IsZero() {
statefields["started_at"] = started.UnixNano() statefields["started_at"] = started.UnixNano()
statefields["uptime_ns"] = finished.Sub(started).Nanoseconds()
uptime := finished.Sub(started)
if finished.Before(started) {
uptime = now().Sub(started)
}
statefields["uptime_ns"] = uptime.Nanoseconds()
} }
acc.AddFields("docker_container_status", statefields, tags, time.Now()) acc.AddFields("docker_container_status", statefields, tags, now())
if info.State.Health != nil { if info.State.Health != nil {
healthfields := map[string]interface{}{ healthfields := map[string]interface{}{
"health_status": info.State.Health.Status, "health_status": info.State.Health.Status,
"failing_streak": info.ContainerJSONBase.State.Health.FailingStreak, "failing_streak": info.ContainerJSONBase.State.Health.FailingStreak,
} }
acc.AddFields("docker_container_health", healthfields, tags, time.Now()) acc.AddFields("docker_container_health", healthfields, tags, now())
} }
} }

198
plugins/inputs/docker/docker_test.go
View File

@ -11,6 +11,7 @@ import (
"github.com/docker/docker/api/types" "github.com/docker/docker/api/types"
"github.com/docker/docker/api/types/swarm" "github.com/docker/docker/api/types/swarm"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/testutil" "github.com/influxdata/telegraf/testutil"
"github.com/stretchr/testify/require" "github.com/stretchr/testify/require"
) )
@ -541,25 +542,22 @@ func TestContainerNames(t *testing.T) {
} }
} }
func TestContainerStatus(t *testing.T) { func FilterMetrics(metrics []telegraf.Metric, f func(telegraf.Metric) bool) []telegraf.Metric {
type expectation struct { results := []telegraf.Metric{}
// tags for _, m := range metrics {
Status string if f(m) {
// fields results = append(results, m)
ContainerID string
OOMKilled bool
Pid int
ExitCode int
StartedAt time.Time
FinishedAt time.Time
UptimeNs int64
} }
}
return results
}
func TestContainerStatus(t *testing.T) {
var tests = []struct { var tests = []struct {
name string name string
now func() time.Time now func() time.Time
inspect types.ContainerJSON inspect types.ContainerJSON
expect expectation expected []telegraf.Metric
}{ }{
{ {
name: "finished_at is zero value", name: "finished_at is zero value",
@ -567,49 +565,141 @@ func TestContainerStatus(t *testing.T) {
return time.Date(2018, 6, 14, 5, 51, 53, 266176036, time.UTC) return time.Date(2018, 6, 14, 5, 51, 53, 266176036, time.UTC)
}, },
inspect: containerInspect(), inspect: containerInspect(),
expect: expectation{ expected: []telegraf.Metric{
ContainerID: "e2173b9478a6ae55e237d4d74f8bbb753f0817192b5081334dc78476296b7dfb", testutil.MustMetric(
Status: "running", "docker_container_status",
OOMKilled: false, map[string]string{
Pid: 1234, "container_name": "etcd",
ExitCode: 0, "container_image": "quay.io/coreos/etcd",
StartedAt: time.Date(2018, 6, 14, 5, 48, 53, 266176036, time.UTC), "container_version": "v2.2.2",
UptimeNs: int64(3 * time.Minute), "engine_host": "absol",
"label1": "test_value_1",
"label2": "test_value_2",
"server_version": "17.09.0-ce",
"container_status": "running",
"source": "e2173b9478a6",
},
map[string]interface{}{
"oomkilled": false,
"pid": 1234,
"exitcode": 0,
"container_id": "e2173b9478a6ae55e237d4d74f8bbb753f0817192b5081334dc78476296b7dfb",
"started_at": time.Date(2018, 6, 14, 5, 48, 53, 266176036, time.UTC).UnixNano(),
"uptime_ns": int64(3 * time.Minute),
},
time.Date(2018, 6, 14, 5, 51, 53, 266176036, time.UTC),
),
}, },
}, },
{ {
name: "finished_at is non-zero value", name: "finished_at is non-zero value",
now: func() time.Time {
return time.Date(2018, 6, 14, 5, 51, 53, 266176036, time.UTC)
},
inspect: func() types.ContainerJSON { inspect: func() types.ContainerJSON {
i := containerInspect() i := containerInspect()
i.ContainerJSONBase.State.FinishedAt = "2018-06-14T05:53:53.266176036Z" i.ContainerJSONBase.State.FinishedAt = "2018-06-14T05:53:53.266176036Z"
return i return i
}(), }(),
expect: expectation{ expected: []telegraf.Metric{
ContainerID: "e2173b9478a6ae55e237d4d74f8bbb753f0817192b5081334dc78476296b7dfb", testutil.MustMetric(
Status: "running", "docker_container_status",
OOMKilled: false, map[string]string{
Pid: 1234, "container_name": "etcd",
ExitCode: 0, "container_image": "quay.io/coreos/etcd",
StartedAt: time.Date(2018, 6, 14, 5, 48, 53, 266176036, time.UTC), "container_version": "v2.2.2",
FinishedAt: time.Date(2018, 6, 14, 5, 53, 53, 266176036, time.UTC), "engine_host": "absol",
UptimeNs: int64(5 * time.Minute), "label1": "test_value_1",
"label2": "test_value_2",
"server_version": "17.09.0-ce",
"container_status": "running",
"source": "e2173b9478a6",
},
map[string]interface{}{
"oomkilled": false,
"pid": 1234,
"exitcode": 0,
"container_id": "e2173b9478a6ae55e237d4d74f8bbb753f0817192b5081334dc78476296b7dfb",
"started_at": time.Date(2018, 6, 14, 5, 48, 53, 266176036, time.UTC).UnixNano(),
"finished_at": time.Date(2018, 6, 14, 5, 53, 53, 266176036, time.UTC).UnixNano(),
"uptime_ns": int64(5 * time.Minute),
},
time.Date(2018, 6, 14, 5, 51, 53, 266176036, time.UTC),
),
}, },
}, },
{ {
name: "started_at is zero value", name: "started_at is zero value",
now: func() time.Time {
return time.Date(2018, 6, 14, 5, 51, 53, 266176036, time.UTC)
},
inspect: func() types.ContainerJSON { inspect: func() types.ContainerJSON {
i := containerInspect() i := containerInspect()
i.ContainerJSONBase.State.StartedAt = "" i.ContainerJSONBase.State.StartedAt = ""
i.ContainerJSONBase.State.FinishedAt = "2018-06-14T05:53:53.266176036Z" i.ContainerJSONBase.State.FinishedAt = "2018-06-14T05:53:53.266176036Z"
return i return i
}(), }(),
expect: expectation{ expected: []telegraf.Metric{
ContainerID: "e2173b9478a6ae55e237d4d74f8bbb753f0817192b5081334dc78476296b7dfb", testutil.MustMetric(
Status: "running", "docker_container_status",
OOMKilled: false, map[string]string{
Pid: 1234, "container_name": "etcd",
ExitCode: 0, "container_image": "quay.io/coreos/etcd",
FinishedAt: time.Date(2018, 6, 14, 5, 53, 53, 266176036, time.UTC), "container_version": "v2.2.2",
"engine_host": "absol",
"label1": "test_value_1",
"label2": "test_value_2",
"server_version": "17.09.0-ce",
"container_status": "running",
"source": "e2173b9478a6",
},
map[string]interface{}{
"oomkilled": false,
"pid": 1234,
"exitcode": 0,
"container_id": "e2173b9478a6ae55e237d4d74f8bbb753f0817192b5081334dc78476296b7dfb",
"finished_at": time.Date(2018, 6, 14, 5, 53, 53, 266176036, time.UTC).UnixNano(),
},
time.Date(2018, 6, 14, 5, 51, 53, 266176036, time.UTC),
),
},
},
{
name: "container has been restarted",
now: func() time.Time {
return time.Date(2019, 1, 1, 0, 0, 3, 0, time.UTC)
},
inspect: func() types.ContainerJSON {
i := containerInspect()
i.ContainerJSONBase.State.StartedAt = "2019-01-01T00:00:02Z"
i.ContainerJSONBase.State.FinishedAt = "2019-01-01T00:00:01Z"
return i
}(),
expected: []telegraf.Metric{
testutil.MustMetric(
"docker_container_status",
map[string]string{
"container_name": "etcd",
"container_image": "quay.io/coreos/etcd",
"container_version": "v2.2.2",
"engine_host": "absol",
"label1": "test_value_1",
"label2": "test_value_2",
"server_version": "17.09.0-ce",
"container_status": "running",
"source": "e2173b9478a6",
},
map[string]interface{}{
"oomkilled": false,
"pid": 1234,
"exitcode": 0,
"container_id": "e2173b9478a6ae55e237d4d74f8bbb753f0817192b5081334dc78476296b7dfb",
"started_at": time.Date(2019, 1, 1, 0, 0, 2, 0, time.UTC).UnixNano(),
"finished_at": time.Date(2019, 1, 1, 0, 0, 1, 0, time.UTC).UnixNano(),
"uptime_ns": int64(1 * time.Second),
},
time.Date(2019, 1, 1, 0, 0, 3, 0, time.UTC),
),
}, },
}, },
} }
@ -643,39 +733,13 @@ func TestContainerStatus(t *testing.T) {
now = time.Now now = time.Now
}() }()
err := acc.GatherError(d.Gather) err := d.Gather(&acc)
require.NoError(t, err) require.NoError(t, err)
fields := map[string]interface{}{ actual := FilterMetrics(acc.GetTelegrafMetrics(), func(m telegraf.Metric) bool {
"oomkilled": tt.expect.OOMKilled, return m.Name() == "docker_container_status"
"pid": tt.expect.Pid,
"exitcode": tt.expect.ExitCode,
"container_id": tt.expect.ContainerID,
}
if started := tt.expect.StartedAt; !started.IsZero() {
fields["started_at"] = started.UnixNano()
fields["uptime_ns"] = tt.expect.UptimeNs
}
if finished := tt.expect.FinishedAt; !finished.IsZero() {
fields["finished_at"] = finished.UnixNano()
}
acc.AssertContainsTaggedFields(t,
"docker_container_status",
fields,
map[string]string{
"container_name": "etcd",
"container_image": "quay.io/coreos/etcd",
"container_version": "v2.2.2",
"engine_host": "absol",
"label1": "test_value_1",
"label2": "test_value_2",
"server_version": "17.09.0-ce",
"container_status": tt.expect.Status,
"source": "e2173b9478a6",
}) })
testutil.RequireMetricsEqual(t, tt.expected, actual)
}) })
} }
} }