Skip to content

feat: Add placement status metrics #1077

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
britaniar wants to merge 19 commits into from
Closed

feat: Add placement status metrics #1077

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
623f26d
add placement status metrics
britaniar Mar 12, 2025
8486603
revert controller_test.go
britaniar Apr 3, 2025
8a89ce8
address comments
britaniar Apr 3, 2025
0200684
try to reduce flakiness
britaniar Apr 3, 2025
c6a8490
only check for complete metric for last scenario
britaniar Apr 3, 2025
ed646ef
address comments
britaniar Apr 4, 2025
a6026c9
update test cases
britaniar Apr 4, 2025
222867e
revert changes
britaniar Apr 4, 2025
f337151
minor change
britaniar Apr 4, 2025
e89bb6e
address comments
britaniar Apr 7, 2025
82a9a65
update test w/ new metric names
britaniar Apr 7, 2025
c74a770
add report diff test and address comments
britaniar Apr 9, 2025
4ed96c6
use cmp.Diff to compare metric and expectedMetric labels
britaniar Apr 9, 2025
d446d00
fix lint errors
britaniar Apr 9, 2025
fef4bb6
update check function & tests
britaniar Apr 10, 2025
07ffce0
move metric comparison to common test utils
britaniar Apr 10, 2025
1cec5f4
fix test
britaniar Apr 10, 2025
9bbdf16
fix test
britaniar Apr 10, 2025
d3e408d
address comments
britaniar Apr 11, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 9 additions & 10 deletions pkg/controllers/clusterresourceplacement/controller.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,8 +94,8 @@ func (r *Reconciler) handleDelete(ctx context.Context, crp *fleetv1beta1.Cluster
return ctrl.Result{}, err
}

metrics.FleetPlacementComplete.DeletePartialMatch(prometheus.Labels{"name": crp.Name})
metrics.FleetPlacementStatus.DeletePartialMatch(prometheus.Labels{"name": crp.Name})
metrics.FleetPlacementCompleteLastTimeStampSeconds.DeletePartialMatch(prometheus.Labels{"name": crp.Name})
metrics.FleetPlacementStatusLastTimeStampSeconds.DeletePartialMatch(prometheus.Labels{"name": crp.Name})
Comment on lines +97 to +98
Copy link
Contributor

@ryanzhang-oss ryanzhang-oss Apr 14, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wonder why move the delete earlier instead of replacing them at line 105?

Copy link
Contributor

@jwtty jwtty Apr 15, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I asked to move these earlier. If the pod stops when finalizer is removed but before metrics are removed, they would be left never deleted.

controllerutil.RemoveFinalizer(crp, fleetv1beta1.ClusterResourcePlacementCleanupFinalizer)
if err := r.Client.Update(ctx, crp); err != nil {
klog.ErrorS(err, "Failed to remove crp finalizer", "clusterResourcePlacement", crpKObj)
Expand Down Expand Up @@ -178,8 +178,8 @@ func (r *Reconciler) handleUpdate(ctx context.Context, crp *fleetv1beta1.Cluster
klog.ErrorS(updateErr, "Failed to update the status", "clusterResourcePlacement", crpKObj)
return ctrl.Result{}, controller.NewUpdateIgnoreConflictError(updateErr)
}
metrics.FleetPlacementStatus.DeletePartialMatch(prometheus.Labels{"name": crp.Name})
metrics.FleetPlacementStatus.WithLabelValues(crp.Name, strconv.FormatInt(crp.Generation, 10), scheduleCondition.Type, string(scheduleCondition.Status)).SetToCurrentTime()
metrics.FleetPlacementStatusLastTimeStampSeconds.DeletePartialMatch(prometheus.Labels{"name": crp.Name})
Copy link
Contributor

@jwtty jwtty Apr 14, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Here you can call emitPlacementStatusMetric too.

metrics.FleetPlacementStatusLastTimeStampSeconds.WithLabelValues(crp.Name, strconv.FormatInt(crp.Generation, 10), scheduleCondition.Type, string(scheduleCondition.Status)).SetToCurrentTime()
return ctrl.Result{}, err
}

Expand Down Expand Up @@ -226,12 +226,12 @@ func (r *Reconciler) handleUpdate(ctx context.Context, crp *fleetv1beta1.Cluster
klog.V(2).InfoS("Placement has finished the rollout process and reached the desired status", "clusterResourcePlacement", crpKObj, "generation", crp.Generation)
r.Recorder.Event(crp, corev1.EventTypeNormal, "PlacementRolloutCompleted", "Placement has finished the rollout process and reached the desired status")
}
metrics.FleetPlacementComplete.WithLabelValues(crp.Name, "true").SetToCurrentTime()
metrics.FleetPlacementCompleteLastTimeStampSeconds.WithLabelValues(crp.Name, "true").SetToCurrentTime()
// We don't need to requeue any request now by watching the binding changes
return ctrl.Result{}, nil
}

metrics.FleetPlacementComplete.WithLabelValues(crp.Name, "false").SetToCurrentTime()
metrics.FleetPlacementCompleteLastTimeStampSeconds.WithLabelValues(crp.Name, "false").SetToCurrentTime()

if !isClusterScheduled {
// Note:
Expand Down Expand Up @@ -1042,27 +1042,26 @@ func isCRPScheduled(crp *fleetv1beta1.ClusterResourcePlacement) bool {
}

func emitPlacementStatusMetric(crp *fleetv1beta1.ClusterResourcePlacement) {
metrics.FleetPlacementStatus.DeletePartialMatch(prometheus.Labels{"name": crp.Name})
// Check CRP Scheduled condition
status := "nil"
cond := crp.GetCondition(string(fleetv1beta1.ClusterResourcePlacementScheduledConditionType))
if !condition.IsConditionStatusTrue(cond, crp.Generation) {
if cond != nil {
status = string(cond.Status)
}
metrics.FleetPlacementStatus.WithLabelValues(crp.Name, strconv.FormatInt(crp.Generation, 10), string(fleetv1beta1.ClusterResourcePlacementScheduledConditionType), status).SetToCurrentTime()
metrics.FleetPlacementStatusLastTimeStampSeconds.WithLabelValues(crp.Name, strconv.FormatInt(crp.Generation, 10), string(fleetv1beta1.ClusterResourcePlacementScheduledConditionType), status).SetToCurrentTime()
return
}

// Check CRP expected conditions
expectedCondTypes := determineExpectedCRPAndResourcePlacementStatusCondType(crp)
for _, condType := range expectedCondTypes {
cond := crp.GetCondition(string(condType.ClusterResourcePlacementConditionType()))
cond = crp.GetCondition(string(condType.ClusterResourcePlacementConditionType()))
if !condition.IsConditionStatusTrue(cond, crp.Generation) {
if cond != nil {
status = string(cond.Status)
}
metrics.FleetPlacementStatus.WithLabelValues(crp.Name, strconv.FormatInt(crp.Generation, 10), string(condType.ClusterResourcePlacementConditionType()), status).SetToCurrentTime()
metrics.FleetPlacementStatusLastTimeStampSeconds.WithLabelValues(crp.Name, strconv.FormatInt(crp.Generation, 10), string(condType.ClusterResourcePlacementConditionType()), status).SetToCurrentTime()
return
}
}
Copy link
Contributor

@zhiying-lin zhiying-lin Apr 11, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There are two options:

  1. two metrics, like the current one, completedMetrics + incompletedPlacementStatus metrics
  2. one metrics, PlacementStatus metrics to record the last condition of CPR. isCompleted = ReportDiff == true || available condition true

Personally i prefer option 2

Copy link
Contributor Author

@britaniar britaniar Apr 11, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I personally prefer option 1. The completed metrics makes it a lot easier to query whether the CRP is completed or not for dashboards. If in the future, not sure if we will, we add more final conditions (like Available and ReportDiff) then we will have to continue changing the alert to include these cases. The second metric was just to provide more information in the alert, while the first helps with dashboard and easier to filter out through the logs.

Copy link
Contributor

@jwtty jwtty Apr 14, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The current status metric never has a "good" status. How are we going to use this metric? I feel this metric is not independent: we always need to use it along with the completed metric, we always need to verify if crp is still incomplete first and then refer to this metric to see why it's still incomplete.

Copy link
Contributor

@zhiying-lin zhiying-lin Apr 14, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

agree with Wantong, they seem coupled, not sure how to join them but with different timestamp.

Copy link
Contributor Author

@britaniar britaniar Apr 15, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Talked to Ryan. We will couple them.

Expand Down
Loading
Loading