enhance: Unify querycoord meta metrics (#38233)

Related to #36456 Unify collection/partition number metrics to collection manager in case of unwant missing modification Signed-off-by: Congqi Xia <[email protected]>
milvus-io · Dec 5, 2024 · 32645fc · 32645fc
1 parent 7944538
commit 32645fc
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 5 deletions.
diff --git a/internal/querycoordv2/job/job_release.go b/internal/querycoordv2/job/job_release.go
@@ -112,8 +112,6 @@ func (job *ReleaseCollectionJob) Execute() error {
 		proxyutil.SetMsgType(commonpb.MsgType_ReleaseCollection))
 
 	waitCollectionReleased(job.dist, job.checkerController, req.GetCollectionID())
-	metrics.QueryCoordNumCollections.WithLabelValues().Dec()
-	metrics.QueryCoordNumPartitions.WithLabelValues().Sub(float64(len(toRelease)))
 	metrics.QueryCoordReleaseCount.WithLabelValues(metrics.TotalLabel).Inc()
 	metrics.QueryCoordReleaseCount.WithLabelValues(metrics.SuccessLabel).Inc()
 	return nil
@@ -196,7 +194,6 @@ func (job *ReleasePartitionJob) Execute() error {
 			log.Warn("failed to remove replicas", zap.Error(err))
 		}
 		job.targetObserver.ReleaseCollection(req.GetCollectionID())
-		metrics.QueryCoordNumCollections.WithLabelValues().Dec()
 		// try best discard cache
 		// shall not affect releasing if failed
 		job.proxyManager.InvalidateCollectionMetaCache(job.ctx,
@@ -216,6 +213,5 @@ func (job *ReleasePartitionJob) Execute() error {
 		job.targetObserver.ReleasePartition(req.GetCollectionID(), toRelease...)
 		waitCollectionReleased(job.dist, job.checkerController, req.GetCollectionID(), toRelease...)
 	}
-	metrics.QueryCoordNumPartitions.WithLabelValues().Sub(float64(len(toRelease)))
 	return nil
 }
diff --git a/internal/querycoordv2/meta/collection_manager.go b/internal/querycoordv2/meta/collection_manager.go
@@ -591,6 +591,7 @@ func (m *CollectionManager) UpdateLoadPercent(ctx context.Context, partitionID i
 		newCollection.RecoverTimes = 0
 
 		metrics.QueryCoordNumCollections.WithLabelValues().Set(float64(len(lo.Values(m.collections))))
+		metrics.QueryCoordNumPartitions.WithLabelValues().Set(float64(len(m.partitions)))
 		elapsed := time.Since(newCollection.CreatedAt)
 		metrics.QueryCoordLoadLatency.WithLabelValues().Observe(float64(elapsed.Milliseconds()))
 		eventlog.Record(eventlog.NewRawEvt(eventlog.Level_Info, fmt.Sprintf("Collection %d loaded", newCollection.CollectionID)))
@@ -616,6 +617,8 @@ func (m *CollectionManager) RemoveCollection(ctx context.Context, collectionID t
 		delete(m.collectionPartitions, collectionID)
 	}
 	metrics.CleanQueryCoordMetricsWithCollectionID(collectionID)
+	metrics.QueryCoordNumPartitions.WithLabelValues().Set(float64(len(m.partitions)))
+	metrics.QueryCoordNumCollections.WithLabelValues().Set(float64(len(lo.Values(m.collections))))
 	return nil
 }
 
@@ -627,7 +630,9 @@ func (m *CollectionManager) RemovePartition(ctx context.Context, collectionID ty
 	m.rwmutex.Lock()
 	defer m.rwmutex.Unlock()
 
-	return m.removePartition(ctx, collectionID, partitionIDs...)
+	err := m.removePartition(ctx, collectionID, partitionIDs...)
+	metrics.QueryCoordNumPartitions.WithLabelValues().Set(float64(len(m.partitions)))
+	return err
 }
 
 func (m *CollectionManager) removePartition(ctx context.Context, collectionID typeutil.UniqueID, partitionIDs ...typeutil.UniqueID) error {