-
Notifications
You must be signed in to change notification settings - Fork 128
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Add AlertBackfill functionality to promxy
This optional config flag will enable promxy to backfill alert status by re-querying the alert statement on startup -- in the event that the downstream store doesn't have the corresponding ALERTS_FOR_STATE series. Fixes #50
- Loading branch information
Showing
12 changed files
with
843 additions
and
11 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
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 was deleted.
Oops, something went wrong.
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 +1,72 @@ | ||
| package alertbackfill | ||
|
|
||
| import ( | ||
| "time" | ||
|
|
||
| "github.com/prometheus/common/model" | ||
| "github.com/prometheus/prometheus/model/labels" | ||
| ) | ||
|
|
||
| // GenerateAlertStateMatrix will generate the a model.Matrix which is the equivalent | ||
| // of the `ALERTS_FOR_STATE` series that would have been generated for the given data | ||
| func GenerateAlertStateMatrix(v model.Matrix, matchers []*labels.Matcher, alertLabels labels.Labels, step time.Duration) model.Matrix { | ||
| matrix := make(model.Matrix, 0, v.Len()) | ||
| MATRIXSAMPLE_LOOP: | ||
| for _, item := range v { | ||
| // clone the metric -- and convert the labels | ||
| metric := item.Metric.Clone() | ||
|
|
||
| // Add the labels which the alert would add | ||
| for _, label := range alertLabels { | ||
| metric[model.LabelName(label.Name)] = model.LabelValue(label.Value) | ||
| } | ||
|
|
||
| // Filter to results that match our matchers | ||
| for _, matcher := range matchers { | ||
| switch matcher.Name { | ||
| // Overwrite the __name__ and alertname | ||
| case model.MetricNameLabel, model.AlertNameLabel: | ||
| metric[model.LabelName(matcher.Name)] = model.LabelValue(matcher.Value) | ||
| default: | ||
| if !matcher.Matches(string(metric[model.LabelName(matcher.Name)])) { | ||
| continue MATRIXSAMPLE_LOOP | ||
| } | ||
| } | ||
| } | ||
|
|
||
| var ( | ||
| activeAt model.SampleValue | ||
| lastPoint time.Time | ||
| ) | ||
| // Now we have to convert the *actual* result into the series that is stored for the ALERTS | ||
| samples := make([]model.SamplePair, len(item.Values)) | ||
| for x, sample := range item.Values { | ||
| sampleTime := sample.Timestamp.Time() | ||
|
|
||
| // If we are missing a point in the matrix; then we are going to assume | ||
| // that the series cleared, so we need to reset activeAt | ||
| if sampleTime.Sub(lastPoint) > step { | ||
| activeAt = 0 | ||
| } | ||
| lastPoint = sampleTime | ||
|
|
||
| // if there is no `activeAt` set; lets set this timestamp (earliest timestamp in the steps that has a point) | ||
| if activeAt == 0 { | ||
| activeAt = model.SampleValue(sample.Timestamp.Unix()) | ||
| } | ||
|
|
||
| samples[x] = model.SamplePair{ | ||
| Timestamp: sample.Timestamp, | ||
| // The timestamp is a unix timestapm of ActiveAt, so we'll set this to the timestamp instead of the value | ||
| Value: activeAt, | ||
| } | ||
| } | ||
|
|
||
| matrix = append(matrix, &model.SampleStream{ | ||
| Metric: metric, | ||
| Values: samples, | ||
| }) | ||
| } | ||
|
|
||
| return matrix | ||
| } |
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 +1,187 @@ | ||
| package alertbackfill | ||
|
|
||
| import ( | ||
| "strconv" | ||
| "testing" | ||
| "time" | ||
|
|
||
| "github.com/prometheus/common/model" | ||
| "github.com/prometheus/prometheus/model/labels" | ||
| ) | ||
|
|
||
| func TestGenerateAlertStateMatrix(t *testing.T) { | ||
| start := model.Time(0).Add(time.Minute) | ||
| tests := []struct { | ||
| in model.Matrix | ||
| matchers []*labels.Matcher | ||
| alertLabels labels.Labels | ||
| step time.Duration | ||
|
|
||
| out model.Matrix | ||
| }{ | ||
| // Simple test case | ||
| { | ||
| in: model.Matrix{ | ||
| &model.SampleStream{ | ||
| Metric: model.Metric{ | ||
| "__name__": "prometheus_build_info", | ||
| "job": "prometheus", | ||
| "replica": "a", | ||
| }, | ||
| Values: []model.SamplePair{ | ||
| {start.Add(time.Minute * 0), 1}, | ||
| {start.Add(time.Minute * 1), 1}, | ||
| {start.Add(time.Minute * 2), 1}, | ||
| {start.Add(time.Minute * 3), 1}, | ||
| {start.Add(time.Minute * 4), 1}, | ||
| // Have a gap! | ||
| {start.Add(time.Minute * 6), 1}, | ||
| }, | ||
| }, | ||
| &model.SampleStream{ | ||
| Metric: model.Metric{ | ||
| "__name__": "prometheus_build_info", | ||
| "job": "prometheus", | ||
| "replica": "b", | ||
| }, | ||
| Values: []model.SamplePair{ | ||
| {start.Add(time.Minute * 0), 1}, | ||
| {start.Add(time.Minute * 1), 1}, | ||
| {start.Add(time.Minute * 2), 1}, | ||
| {start.Add(time.Minute * 3), 1}, | ||
| {start.Add(time.Minute * 4), 1}, | ||
| }, | ||
| }, | ||
| }, | ||
| matchers: []*labels.Matcher{ | ||
| labels.MustNewMatcher(labels.MatchEqual, model.MetricNameLabel, "ALERTS_FOR_STATE"), | ||
| labels.MustNewMatcher(labels.MatchEqual, model.AlertNameLabel, "testalert"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "replica", "a"), | ||
| }, | ||
| alertLabels: labels.Labels{ | ||
| labels.Label{"severity", "page"}, | ||
| }, | ||
| step: time.Minute, | ||
| out: model.Matrix{ | ||
| &model.SampleStream{ | ||
| Metric: model.Metric{ | ||
| "__name__": "ALERTS_FOR_STATE", | ||
| "alertname": "testalert", | ||
| "job": "prometheus", | ||
| "replica": "a", | ||
| "severity": "page", | ||
| }, | ||
| Values: []model.SamplePair{ | ||
| {start.Add(time.Minute * 0), model.SampleValue(start.Unix())}, | ||
| {start.Add(time.Minute * 1), model.SampleValue(start.Unix())}, | ||
| {start.Add(time.Minute * 2), model.SampleValue(start.Unix())}, | ||
| {start.Add(time.Minute * 3), model.SampleValue(start.Unix())}, | ||
| {start.Add(time.Minute * 4), model.SampleValue(start.Unix())}, | ||
| {start.Add(time.Minute * 6), model.SampleValue(start.Add(time.Minute * 6).Unix())}, | ||
| }, | ||
| }, | ||
| }, | ||
| }, | ||
|
|
||
| // Example from `prometheus_build_info == 1` | ||
| { | ||
| in: model.Matrix{ | ||
| &model.SampleStream{ | ||
| Metric: model.Metric{ | ||
| "__name__": "prometheus_build_info", | ||
| "branch": "HEAD", | ||
| "goversion": "go1.16.4", | ||
| "instance": "demo.do.prometheus.io:9090", | ||
| "job": "prometheus", | ||
| "replica": "a", | ||
| "revision": "24c9b61221f7006e87cd62b9fe2901d43e19ed53", | ||
| "version": "2.27.0", | ||
| }, | ||
| Values: []model.SamplePair{ | ||
| {model.TimeFromUnix(1692419110), 1}, | ||
| {model.TimeFromUnix(1692419115), 1}, | ||
| {model.TimeFromUnix(1692419120), 1}, | ||
| {model.TimeFromUnix(1692419125), 1}, | ||
| {model.TimeFromUnix(1692419130), 1}, | ||
| }, | ||
| }, | ||
| &model.SampleStream{ | ||
| Metric: model.Metric{ | ||
| "__name__": "prometheus_build_info", | ||
| "branch": "HEAD", | ||
| "goversion": "go1.16.4", | ||
| "instance": "demo.do.prometheus.io:9090", | ||
| "job": "prometheus", | ||
| "replica": "b", | ||
| "revision": "24c9b61221f7006e87cd62b9fe2901d43e19ed53", | ||
| "version": "2.27.0", | ||
| }, | ||
| Values: []model.SamplePair{ | ||
| {model.TimeFromUnix(1692419110), 1}, | ||
| {model.TimeFromUnix(1692419115), 1}, | ||
| {model.TimeFromUnix(1692419120), 1}, | ||
| {model.TimeFromUnix(1692419125), 1}, | ||
| {model.TimeFromUnix(1692419130), 1}, | ||
| }, | ||
| }, | ||
| }, | ||
| matchers: []*labels.Matcher{ | ||
| labels.MustNewMatcher(labels.MatchEqual, model.MetricNameLabel, "ALERTS_FOR_STATE"), | ||
| labels.MustNewMatcher(labels.MatchEqual, model.AlertNameLabel, "testAlert"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "branch", "HEAD"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "goversion", "go1.16.4"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "instance", "demo.do.prometheus.io:9090"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "job", "prometheus"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "replica", "a"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "revision", "24c9b61221f7006e87cd62b9fe2901d43e19ed53"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "severity", "pageMORE"), | ||
| labels.MustNewMatcher(labels.MatchEqual, "version", "2.27.0"), | ||
| }, | ||
| alertLabels: labels.Labels{ | ||
| labels.Label{"severity", "pageMORE"}, | ||
| }, | ||
| step: time.Second * 5, | ||
| out: model.Matrix{ | ||
| &model.SampleStream{ | ||
| Metric: model.Metric{ | ||
| "__name__": "ALERTS_FOR_STATE", | ||
| "alertname": "testAlert", | ||
| "branch": "HEAD", | ||
| "goversion": "go1.16.4", | ||
| "instance": "demo.do.prometheus.io:9090", | ||
| "job": "prometheus", | ||
| "replica": "a", | ||
| "revision": "24c9b61221f7006e87cd62b9fe2901d43e19ed53", | ||
| "version": "2.27.0", | ||
| "severity": "pageMORE", | ||
| }, | ||
| Values: []model.SamplePair{ | ||
| {model.TimeFromUnix(1692419110), 1692419110}, | ||
| {model.TimeFromUnix(1692419115), 1692419110}, | ||
| {model.TimeFromUnix(1692419120), 1692419110}, | ||
| {model.TimeFromUnix(1692419125), 1692419110}, | ||
| {model.TimeFromUnix(1692419130), 1692419110}, | ||
| }, | ||
| }, | ||
| }, | ||
| }, | ||
| } | ||
|
|
||
| for i, test := range tests { | ||
| t.Run(strconv.Itoa(i), func(t *testing.T) { | ||
| out := GenerateAlertStateMatrix(test.in, test.matchers, test.alertLabels, test.step) | ||
|
|
||
| if test.out.String() != out.String() { | ||
| t.Fatalf("mismatch in series expected=%v actual=%v", test.out, out) | ||
| } | ||
|
|
||
| for i, sampleStream := range out { | ||
| for _, matcher := range test.matchers { | ||
| if !matcher.Matches(string(sampleStream.Metric[model.LabelName(matcher.Name)])) { | ||
| t.Fatalf("out series=%d label %s=%s doesn't match matcher %v", i, matcher.Name, sampleStream.Metric[model.LabelName(matcher.Name)], matcher.Value) | ||
| } | ||
| } | ||
| } | ||
| }) | ||
| } | ||
| } |
Oops, something went wrong.