opensource/cloudpods
3
0
Fork 0
mirror of https://github.com/yunionio/cloudpods.git synced 2026-10-24 12:32:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #14965 from zexi/automated-cherry-pick-of-#14964-upstream-master

Browse Source 
Automated cherry pick of #14964: fix(monitor): alert resource not recorverd
This commit is contained in:
Zexi Li
2022-09-14 10:05:41 +08:00
committed by GitHub
parent c50b494d01 ce27f26404
commit 4d70b3d5ec
3 changed files with 29 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pkg/monitor/alerting/conditions/query.go
+1 -1
View File
@@ -178,7 +178,7 @@ func (c *QueryCondition) Eval(context *alerting.EvalContext) (*alerting.Conditio
if context.IsTestRun {
context.Logs = append(context.Logs, &monitor.ResultLogEntry{
Message: fmt.Sprintf("Condition[%d]: Eval: %v, Metric: %s, Value: %v", c.Index, evalMatch, series.Name, reducedValue),
Message: fmt.Sprintf("Condition[%d]: EvalMatch: %v, Metric: %s, Value: %s", c.Index, evalMatch, series.Name, jsonutils.Marshal(reducedValue)),
})
}
pkg/monitor/models/alert.go
+2 -2
View File
@@ -615,10 +615,10 @@ func (alert *SAlert) PerformTestRun(
query jsonutils.JSONObject,
input monitor.AlertTestRunInput,
) (*monitor.AlertTestRunOutput, error) {
return alert.testRunAlert(userCred, input)
return alert.TestRunAlert(userCred, input)
}
func (alert *SAlert) testRunAlert(userCred mcclient.TokenCredential, input monitor.AlertTestRunInput) (*monitor.AlertTestRunOutput, error) {
func (alert *SAlert) TestRunAlert(userCred mcclient.TokenCredential, input monitor.AlertTestRunInput) (*monitor.AlertTestRunOutput, error) {
return AlertManager.GetTester().DoTest(alert, userCred, input)
}
pkg/monitor/models/commonalert.go
+26 -31
View File
@@ -1324,11 +1324,8 @@ func (alert *SCommonAlert) StartUpdateMonitorAlertJointTask(ctx context.Context,
}
func (alert *SCommonAlert) UpdateMonitorResourceJoint(ctx context.Context, userCred mcclient.TokenCredential) error {
setting, _ := alert.GetSettings()
inputQuery := monitor.MetricInputQuery{
MetricQuery: make([]*monitor.AlertQuery, 0),
}
var resType string
setting, _ := alert.GetSettings()
for _, con := range setting.Conditions {
measurement, _ := MetricMeasurementManager.GetCache().Get(con.Query.Model.Measurement)
if measurement == nil {
@@ -1336,24 +1333,15 @@ func (alert *SCommonAlert) UpdateMonitorResourceJoint(ctx context.Context, userC
} else {
resType = measurement.ResType
}
//if !utils.IsInStringArray(measurement.ResType, []string{monitor.METRIC_RES_TYPE_HOST,
// monitor.METRIC_RES_TYPE_GUEST, monitor.METRIC_RES_TYPE_AGENT}) {
// continue
//}
inputQuery.MetricQuery = append(inputQuery.MetricQuery, &con.Query)
}
if len(inputQuery.MetricQuery) == 0 {
return nil
}
metrics, err := doQuery(inputQuery)
ret, err := alert.TestRunAlert(userCred, monitor.AlertTestRunInput{})
if err != nil {
return errors.Wrap(err, "doQuery err")
return errors.Wrapf(err, "TestRunAlert %s", alert.GetName())
}
resourceIds := make([]string, 0)
for _, serie := range metrics.Series {
for _, em := range ret.EvalMatches {
resourceKeyId := monitor.MEASUREMENT_TAG_ID[resType]
resourceId := serie.Tags[resourceKeyId]
resourceId := em.Tags[resourceKeyId]
if len(resourceId) == 0 {
continue
}
@@ -1376,27 +1364,34 @@ jointLoop:
}
deleteJointIds = append(deleteJointIds, joint.RowId)
}
if len(resourceIds) == 0 {
if len(resourceIds) == 0 && len(deleteJointIds) == 0 {
return nil
}
if len(deleteJointIds) != 0 {
// sync joints should be deleted
if len(deleteJointIds) > 0 {
err := MonitorResourceAlertManager.DetachJoint(ctx, userCred, monitor.MonitorResourceJointListInput{JointId: deleteJointIds})
if err != nil {
return errors.Wrapf(err, "DetachJoint By alertName:%s err", alert.GetName())
return errors.Wrapf(err, "DetachJoint by alert %s(%s)", alert.GetName(), alert.GetId())
}
}
monitorResources, _ := MonitorResourceManager.GetMonitorResources(monitor.MonitorResourceListInput{ResId: resourceIds})
errs := make([]error, 0)
for _, monRes := range monitorResources {
err := monRes.AttachAlert(ctx, userCred, alert.GetId())
if err != nil {
errs = append(errs, err)
}
monRes.UpdateAlertState()
}
if len(errs) != 0 {
return errors.NewAggregate(errs)
if len(resourceIds) > 0 {
monitorResources, _ := MonitorResourceManager.GetMonitorResources(monitor.MonitorResourceListInput{ResId: resourceIds})
errs := make([]error, 0)
for _, monRes := range monitorResources {
resDesc := fmt.Sprintf("%s/%s/%s", monRes.ResType, monRes.GetName(), monRes.ResId)
if err := monRes.AttachAlert(ctx, userCred, alert.GetId()); err != nil {
errs = append(errs, errors.Wrapf(err, "AttachAlert %s to %s", alert.GetName(), resDesc))
}
if err := monRes.UpdateAlertState(); err != nil {
errs = append(errs, errors.Wrapf(err, "UpdateAlertState for monitor resource %s", resDesc))
}
}
if len(errs) != 0 {
return errors.NewAggregate(errs)
}
}
return nil
}