forked from kubeflow/kubeflow
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
feat(centraldashboard): Displaying metrics from Prometheus (kubeflow#…
…7116) * # backend Adding /api/metrics endpoint for `resourceChartsLink` and `resourceChartsLinkText` field. Adding PrometheusMetricsService class implementing MetricsService Adding `prometheus-query` package ## MetricsService Adding `getChartsLink(): MetricsInfo` to MetricsService. It will be used to send client the info about the metrics (resourceChartsLink and resourceChartsLinkText). ## env - PROMETHEUS_URL: url for prometheus query - METRICS_DASHBOARD: if defined, will add a button under graphs to go to the url defined ## Prometheus query - getNodeCpuUtilization: `sum(rate(node_cpu_seconds_total[5m])) by (instance)` - getPodCpuUtilization: `sum(rate(container_cpu_usage_seconds_total[5m]))` - getPodMemoryUsage: `sum(container_memory_usage_bytes)` # frontend Adding iron-ajax to `/api/metrics`. Adding `metrics` properties to main-page and dashboard-view. Making footer of `ressource-chart.js` hidden if resourceChartsLink is undefined. * Fixing package-lock.json * Adding Unit Tests and Improvements
- Loading branch information
1 parent
e3e3f2b
commit 0fc20c3
Showing
16 changed files
with
340 additions
and
27 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
90 changes: 90 additions & 0 deletions
90
components/centraldashboard/app/prometheus_metrics_service.ts
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,90 @@ | ||
import {Interval, MetricsInfo, MetricsService, TimeSeriesPoint} from "./metrics_service"; | ||
import {PrometheusDriver, RangeVector, ResponseType} from 'prometheus-query'; | ||
|
||
export class PrometheusMetricsService implements MetricsService { | ||
private readonly prometheusDriver: PrometheusDriver; | ||
private readonly dashboardUrl: string | undefined; | ||
|
||
constructor(prometheusDriver: PrometheusDriver, dashboardUrl: string | undefined) { | ||
this.prometheusDriver = prometheusDriver; | ||
this.dashboardUrl = dashboardUrl; | ||
} | ||
|
||
async getNodeCpuUtilization(interval: Interval): Promise<TimeSeriesPoint[]> { | ||
const query = `sum(rate(node_cpu_seconds_total[5m])) by (instance)`; | ||
const result = await this.queryPrometheus(query, this.getCorrespondingTime(interval)); | ||
return this.convertToTimeSeriesPoints(result); | ||
} | ||
|
||
async getPodCpuUtilization(interval: Interval): Promise<TimeSeriesPoint[]> { | ||
const query = `sum(rate(container_cpu_usage_seconds_total[5m]))`; | ||
const result = await this.queryPrometheus(query, this.getCorrespondingTime(interval)); | ||
return this.convertToTimeSeriesPoints(result); | ||
} | ||
|
||
async getPodMemoryUsage(interval: Interval): Promise<TimeSeriesPoint[]> { | ||
const query = `sum(container_memory_usage_bytes)`; | ||
const result = await this.queryPrometheus(query, this.getCorrespondingTime(interval)); | ||
return this.convertToTimeSeriesPoints(result); | ||
} | ||
|
||
private async queryPrometheus(query: string, start: number, end: number = Date.now()): Promise<RangeVector[]> { | ||
const result = await this.prometheusDriver.rangeQuery(query, start, end, 10); | ||
if(result.resultType !== ResponseType.MATRIX) { | ||
console.warn(`The prometheus server returned invalid result type: ${result.resultType}`); | ||
return []; | ||
} | ||
return result.result as RangeVector[]; | ||
} | ||
|
||
private getCorrespondingTime(interval: Interval): number { | ||
let minutes = 0; | ||
switch (interval) { | ||
case Interval.Last5m: | ||
minutes = 5; | ||
break; | ||
case Interval.Last15m: | ||
minutes = 15; | ||
break; | ||
case Interval.Last30m: | ||
minutes = 30; | ||
break; | ||
case Interval.Last60m: | ||
minutes = 60; | ||
break; | ||
case Interval.Last180m: | ||
minutes = 180; | ||
break; | ||
default: | ||
console.warn("unknown interval."); | ||
} | ||
return Date.now() - minutes * 60 * 1000; | ||
} | ||
|
||
private convertToTimeSeriesPoints(series: RangeVector[]): TimeSeriesPoint[] { | ||
const timeSeriesPoints: TimeSeriesPoint[] = []; | ||
series.forEach(serie => { | ||
|
||
const label = Object.entries(serie.metric.labels).map((entry) => { | ||
return entry[0] + "=" + entry[1]; | ||
}).join(","); | ||
|
||
// The `public/components/resource-chart.js` is multiplying the timestamp by 1000 and the value by 100 | ||
serie.values.forEach(value => { | ||
timeSeriesPoints.push({ | ||
timestamp: value.time.getTime() / 1000, | ||
label, | ||
value: value.value / 100, | ||
}); | ||
}); | ||
}); | ||
return timeSeriesPoints; | ||
} | ||
|
||
getChartsLink(): MetricsInfo { | ||
return { | ||
resourceChartsLink: this.dashboardUrl, | ||
resourceChartsLinkText: 'View in dashboard' | ||
}; | ||
} | ||
} |
142 changes: 142 additions & 0 deletions
142
components/centraldashboard/app/prometheus_metrics_service_test.ts
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,142 @@ | ||
import {Metric, PrometheusDriver, QueryResult, ResponseType} from "prometheus-query"; | ||
import {PrometheusMetricsService} from "./prometheus_metrics_service"; | ||
import {Interval, MetricsService, TimeSeriesPoint} from "./metrics_service"; | ||
import {SampleValue} from "prometheus-query/dist/types"; | ||
|
||
type MetricsServiceKeys = keyof MetricsService; | ||
const methods: MetricsServiceKeys[] = ["getNodeCpuUtilization", "getPodCpuUtilization", "getPodMemoryUsage"]; | ||
const queries: {[id: string]: string} = { | ||
"getNodeCpuUtilization": "sum(rate(node_cpu_seconds_total[5m])) by (instance)", | ||
"getPodCpuUtilization": "sum(rate(container_cpu_usage_seconds_total[5m]))", | ||
"getPodMemoryUsage": "sum(container_memory_usage_bytes)" | ||
}; | ||
|
||
const fixedDate = 1557705600000; | ||
|
||
const emptyDataSet: QueryResult = {"resultType": ResponseType.MATRIX,"result":[]}; | ||
const singleInstanceDataSet: QueryResult = { | ||
"resultType": ResponseType.MATRIX, | ||
"result":[ | ||
{ | ||
"metric": {"labels": {"instance":"one"}} as Metric, | ||
"values":[ | ||
{ | ||
time: new Date(fixedDate), | ||
value: 95.5, | ||
} as SampleValue | ||
] | ||
} | ||
] | ||
}; | ||
const multipleInstancesDataSet: QueryResult = { | ||
"resultType": ResponseType.MATRIX, | ||
"result":[ | ||
{ | ||
"metric": {"labels": {"instance":"one"}} as Metric, | ||
"values":[ | ||
{ | ||
time: new Date(fixedDate), | ||
value: 1.0, | ||
} as SampleValue | ||
] | ||
}, | ||
{ | ||
"metric": {"labels": {"instance":"two"}} as Metric, | ||
"values":[ | ||
{ | ||
time: new Date(fixedDate), | ||
value: 2.0, | ||
} as SampleValue | ||
] | ||
}, | ||
{ | ||
"metric": {"labels": {"instance":"three"}} as Metric, | ||
"values":[ | ||
{ | ||
time: new Date(fixedDate), | ||
value: 3.0, | ||
} as SampleValue | ||
] | ||
} | ||
] | ||
}; | ||
|
||
describe('PrometheusMetricsService', () => { | ||
let prometheusDriverClient: jasmine.SpyObj<PrometheusDriver>; | ||
let service: PrometheusMetricsService; | ||
|
||
beforeEach(() => { | ||
jasmine.clock().install(); | ||
jasmine.clock().mockDate(new Date(1557705600000)); | ||
prometheusDriverClient = jasmine.createSpyObj<PrometheusDriver>( | ||
'prometheusDriverClient', ['rangeQuery']); | ||
|
||
service = | ||
new PrometheusMetricsService(prometheusDriverClient, undefined); | ||
}); | ||
|
||
// Iterate over all methods since they have the same behavior | ||
methods.forEach((method) => { | ||
describe(method, async () => { | ||
it('Empty return', async () => { | ||
prometheusDriverClient.rangeQuery.withArgs( | ||
queries[method], | ||
Date.now() - 5 * 60 * 1000, | ||
Date.now(), | ||
10 | ||
).and.returnValue(Promise.resolve(emptyDataSet)); | ||
|
||
const emptyResult = await service[method](Interval.Last5m); | ||
expect(emptyResult).toEqual(Array.of<TimeSeriesPoint>()); | ||
}); | ||
|
||
it('One instance', async () => { | ||
prometheusDriverClient.rangeQuery.withArgs( | ||
queries[method], | ||
Date.now() - 5 * 60 * 1000, | ||
Date.now(), | ||
10 | ||
).and.returnValue(Promise.resolve(singleInstanceDataSet)); | ||
|
||
const singleInstanceResult = await service[method](Interval.Last5m); | ||
expect(singleInstanceResult).toEqual(Array.of<TimeSeriesPoint>({ | ||
timestamp: fixedDate / 1000, | ||
value: 0.955, | ||
label: "instance=one" | ||
})); | ||
}); | ||
|
||
it('Multiple instances', async () => { | ||
prometheusDriverClient.rangeQuery.withArgs( | ||
queries[method], | ||
Date.now() - 5 * 60 * 1000, | ||
Date.now(), | ||
10 | ||
).and.returnValue(Promise.resolve(multipleInstancesDataSet)); | ||
|
||
const singleInstanceResult = await service[method](Interval.Last5m); | ||
expect(singleInstanceResult).toEqual( | ||
Array.of<TimeSeriesPoint>({ | ||
timestamp: fixedDate / 1000, | ||
value: 0.010, | ||
label: "instance=one" | ||
}, | ||
{ | ||
timestamp: fixedDate / 1000, | ||
value: 0.020, | ||
label: "instance=two" | ||
}, | ||
{ | ||
timestamp: fixedDate / 1000, | ||
value: 0.030, | ||
label: "instance=three" | ||
}) | ||
); | ||
}); | ||
}); | ||
}); | ||
|
||
afterEach(() => { | ||
jasmine.clock().uninstall(); | ||
}); | ||
}); |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.