CL-Softwares
/
telegraf
8
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat: add nomad input plugin (#10106)

This commit is contained in:
Francesco Bartolini 2021-12-08 17:34:11 +01:00 committed by GitHub
parent 526dcb4bc8
commit dbf55535c7
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
6 changed files with 411 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
plugins/inputs/all/all.go
View File

@ -128,6 +128,7 @@ import (
_ "github.com/influxdata/telegraf/plugins/inputs/nginx_sts"
_ "github.com/influxdata/telegraf/plugins/inputs/nginx_upstream_check"
_ "github.com/influxdata/telegraf/plugins/inputs/nginx_vts"
_ "github.com/influxdata/telegraf/plugins/inputs/nomad"
_ "github.com/influxdata/telegraf/plugins/inputs/nsd"
_ "github.com/influxdata/telegraf/plugins/inputs/nsq"
_ "github.com/influxdata/telegraf/plugins/inputs/nsq_consumer"

28
plugins/inputs/nomad/README.md Normal file
View File

@ -0,0 +1,28 @@
# Hashicorp Nomad Input Plugin
The Nomad plugin must grab metrics from every Nomad agent of the cluster. Telegraf may be present in every node and connect to the agent locally. In this case should be something like `http://127.0.0.1:4646`.
> Tested on Nomad 1.1.6
## Configuration
```toml
[[inputs.nomad]]
## URL for the Nomad agent
# url = "http://127.0.0.1:4646"
## Set response_timeout (default 5 seconds)
# response_timeout = "5s"
## Optional TLS Config
# tls_ca = /path/to/cafile
# tls_cert = /path/to/certfile
# tls_key = /path/to/keyfile
```
## Metrics
Both Nomad servers and agents collect various metrics. For every details, please have a look at Nomad following documentation:
- [https://www.nomadproject.io/docs/operations/metrics](https://www.nomadproject.io/docs/operations/metrics)
- [https://www.nomadproject.io/docs/operations/telemetry](https://www.nomadproject.io/docs/operations/telemetry)

177
plugins/inputs/nomad/nomad.go Normal file
View File

@ -0,0 +1,177 @@
package nomad
import (
"encoding/json"
"fmt"
"net/http"
"time"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/config"
"github.com/influxdata/telegraf/plugins/common/tls"
"github.com/influxdata/telegraf/plugins/inputs"
)
// Nomad configuration object
type Nomad struct {
URL string `toml:"url"`
ResponseTimeout config.Duration `toml:"response_timeout"`
tls.ClientConfig
roundTripper http.RoundTripper
}
const timeLayout = "2006-01-02 15:04:05 -0700 MST"
var sampleConfig = `
## URL for the Nomad agent
# url = "http://127.0.0.1:4646"
## Set response_timeout (default 5 seconds)
# response_timeout = "5s"
## Optional TLS Config
# tls_ca = /path/to/cafile
# tls_cert = /path/to/certfile
# tls_key = /path/to/keyfile
`
func init() {
inputs.Add("nomad", func() telegraf.Input {
return &Nomad{
ResponseTimeout: config.Duration(5 * time.Second),
}
})
}
// SampleConfig returns a sample config
func (n *Nomad) SampleConfig() string {
return sampleConfig
}
// Description returns a description of the plugin
func (n *Nomad) Description() string {
return "Read metrics from the Nomad API"
}
func (n *Nomad) Init() error {
if n.URL == "" {
n.URL = "http://127.0.0.1:4646"
}
tlsCfg, err := n.ClientConfig.TLSConfig()
if err != nil {
return fmt.Errorf("setting up TLS configuration failed: %v", err)
}
n.roundTripper = &http.Transport{
TLSHandshakeTimeout: 5 * time.Second,
TLSClientConfig: tlsCfg,
ResponseHeaderTimeout: time.Duration(n.ResponseTimeout),
}
return nil
}
// Gather, collects metrics from Nomad endpoint
func (n *Nomad) Gather(acc telegraf.Accumulator) error {
summaryMetrics := &MetricsSummary{}
err := n.loadJSON(n.URL+"/v1/metrics", summaryMetrics)
if err != nil {
return err
}
err = buildNomadMetrics(acc, summaryMetrics)
if err != nil {
return err
}
return nil
}
func (n *Nomad) loadJSON(url string, v interface{}) error {
req, err := http.NewRequest("GET", url, nil)
if err != nil {
return err
}
resp, err := n.roundTripper.RoundTrip(req)
if err != nil {
return fmt.Errorf("error making HTTP request to %s: %s", url, err)
}
defer resp.Body.Close()
if resp.StatusCode != http.StatusOK {
return fmt.Errorf("%s returned HTTP status %s", url, resp.Status)
}
err = json.NewDecoder(resp.Body).Decode(v)
if err != nil {
return fmt.Errorf("error parsing json response: %s", err)
}
return nil
}
// buildNomadMetrics, it builds all the metrics and adds them to the accumulator)
func buildNomadMetrics(acc telegraf.Accumulator, summaryMetrics *MetricsSummary) error {
t, err := time.Parse(timeLayout, summaryMetrics.Timestamp)
if err != nil {
return fmt.Errorf("error parsing time: %s", err)
}
for _, counters := range summaryMetrics.Counters {
tags := counters.DisplayLabels
fields := map[string]interface{}{
"count": counters.Count,
"rate": counters.Rate,
"sum": counters.Sum,
"sumsq": counters.SumSq,
"min": counters.Min,
"max": counters.Max,
"mean": counters.Mean,
}
acc.AddCounter(counters.Name, fields, tags, t)
}
for _, gauges := range summaryMetrics.Gauges {
tags := gauges.DisplayLabels
fields := map[string]interface{}{
"value": gauges.Value,
}
acc.AddGauge(gauges.Name, fields, tags, t)
}
for _, points := range summaryMetrics.Points {
tags := make(map[string]string)
fields := map[string]interface{}{
"value": points.Points,
}
acc.AddFields(points.Name, fields, tags, t)
}
for _, samples := range summaryMetrics.Samples {
tags := samples.DisplayLabels
fields := map[string]interface{}{
"count": samples.Count,
"rate": samples.Rate,
"sum": samples.Sum,
"stddev": samples.Stddev,
"sumsq": samples.SumSq,
"min": samples.Min,
"max": samples.Max,
"mean": samples.Mean,
}
acc.AddCounter(samples.Name, fields, tags, t)
}
return nil
}

53
plugins/inputs/nomad/nomad_metrics.go Normal file
View File

@ -0,0 +1,53 @@
package nomad
import (
"time"
)
type MetricsSummary struct {
Timestamp string `json:"timestamp"`
Gauges []GaugeValue `json:"gauges"`
Points []PointValue `json:"points"`
Counters []SampledValue `json:"counters"`
Samples []SampledValue `json:"samples"`
}
type GaugeValue struct {
Name string `json:"name"`
Hash string `json:"-"`
Value float32 `json:"value"`
Labels []Label `json:"-"`
DisplayLabels map[string]string `json:"Labels"`
}
type PointValue struct {
Name string `json:"name"`
Points []float32 `json:"points"`
}
type SampledValue struct {
Name string `json:"name"`
Hash string `json:"-"`
*AggregateSample
Mean float64 `json:"mean"`
Stddev float64 `json:"stddev"`
Labels []Label `json:"-"`
DisplayLabels map[string]string `json:"Labels"`
}
type AggregateSample struct {
Count int `json:"count"`
Rate float64 `json:"rate"`
Sum float64 `json:"sum"`
SumSq float64 `json:"-"`
Min float64 `json:"min"`
Max float64 `json:"max"`
LastUpdated time.Time `json:"-"`
}
type Label struct {
Name string `json:"name"`
Value string `json:"value"`
}

104
plugins/inputs/nomad/nomad_test.go Normal file
View File

@ -0,0 +1,104 @@
package nomad
import (
"fmt"
"io/ioutil"
"net/http"
"net/http/httptest"
"testing"
"time"
"github.com/influxdata/telegraf"
"github.com/influxdata/telegraf/testutil"
"github.com/stretchr/testify/require"
)
func TestNomadStats(t *testing.T) {
var applyTests = []struct {
name string
expected []telegraf.Metric
}{
{
name: "Metrics",
expected: []telegraf.Metric{
testutil.MustMetric(
"nomad.nomad.rpc.query",
map[string]string{
"host": "node1",
},
map[string]interface{}{
"count": int(7),
"max": float64(1),
"min": float64(1),
"mean": float64(1),
"rate": float64(0.7),
"sum": float64(7),
"sumsq": float64(0),
},
time.Unix(1636843140, 0),
1,
),
testutil.MustMetric(
"nomad.client.allocated.cpu",
map[string]string{
"node_scheduling_eligibility": "eligible",
"host": "node1",
"node_id": "2bbff078-8473-a9de-6c5e-42b4e053e12f",
"datacenter": "dc1",
"node_class": "none",
"node_status": "ready",
},
map[string]interface{}{
"value": float32(500),
},
time.Unix(1636843140, 0),
2,
),
testutil.MustMetric(
"nomad.memberlist.gossip",
map[string]string{
"host": "node1",
},
map[string]interface{}{
"count": int(20),
"max": float64(0.03747599944472313),
"mean": float64(0.013159099989570678),
"min": float64(0.003459000028669834),
"rate": float64(0.026318199979141355),
"stddev": float64(0.009523742715522742),
"sum": float64(0.26318199979141355),
"sumsq": float64(0),
},
time.Unix(1636843140, 0),
1,
),
},
},
}
for _, tt := range applyTests {
t.Run(tt.name, func(t *testing.T) {
ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.RequestURI == "/v1/metrics" {
w.WriteHeader(http.StatusOK)
responseKeyMetrics, _ := ioutil.ReadFile("testdata/response_key_metrics.json")
_, err := fmt.Fprintln(w, string(responseKeyMetrics))
require.NoError(t, err)
}
}))
defer ts.Close()
plugin := &Nomad{
URL: ts.URL,
}
err := plugin.Init()
require.NoError(t, err)
acc := testutil.Accumulator{}
err = plugin.Gather(&acc)
require.NoError(t, err)
testutil.RequireMetricsEqual(t, tt.expected, acc.GetTelegrafMetrics())
})
}
}

48
plugins/inputs/nomad/testdata/response_key_metrics.json vendored Normal file
View File

@ -0,0 +1,48 @@
{
"Counters": [
{
"Count": 7,
"Labels": {
"host": "node1"
},
"Max": 1,
"Mean": 1,
"Min": 1,
"Name": "nomad.nomad.rpc.query",
"Rate": 0.7,
"Stddev": 0,
"Sum": 7
}
],
"Gauges": [
{
"Labels": {
"node_scheduling_eligibility": "eligible",
"host": "node1",
"node_id": "2bbff078-8473-a9de-6c5e-42b4e053e12f",
"datacenter": "dc1",
"node_class": "none",
"node_status": "ready"
},
"Name": "nomad.client.allocated.cpu",
"Value": 500
}
],
"Points": [],
"Samples": [
{
"Count": 20,
"Labels": {
"host": "node1"
},
"Max": 0.03747599944472313,
"Mean": 0.013159099989570678,
"Min": 0.003459000028669834,
"Name": "nomad.memberlist.gossip",
"Rate": 0.026318199979141355,
"Stddev": 0.009523742715522742,
"Sum": 0.26318199979141355
}
],
"Timestamp": "2021-11-13 22:39:00 +0000 UTC"
}