Flink metrics-Managed Service for Prometheus(Prometheus)-阿里云帮助中心

Managed Service for Prometheus collects metrics from Alibaba Cloud Realtime Compute for Apache Flink to help you monitor job health, throughput, latency, checkpoint performance, and JVM resource usage.

Flink metrics follow a hierarchical naming pattern that reflects where each metric is collected:

flink_jobmanager_* -- Cluster-level metrics reported by the JobManager.
flink_taskmanager_job_task_* -- Task-level metrics reported by individual TaskManagers.
flink_taskmanager_job_task_operator_* -- Operator-level metrics scoped to a specific operator within a task.

Use these prefixes to target the right scope when writing PromQL queries or configuring alert rules.

Important

Managed Service for Prometheus charges based on the volume of data written or the number of data points reported. Metrics fall into two categories:

Basic metrics -- Reporting basic metrics from Alibaba Cloud Realtime Compute for Apache Flink to Managed Service for Prometheus is free of charge. This benefit does not apply to other Flink services, such as self-managed clusters.
Custom metrics -- All other metrics are custom metrics. Custom metrics have been billed since January 6, 2020.

Job health

Metric	Description	Details	Unit	Type
`flink_jobmanager_job_numRestarts`	Number of job restarts caused by errors	Counts error-triggered restarts only. Restarts from JobManager failovers are excluded. A steadily increasing value indicates recurring failures that require investigation.	Count	Custom
`flink_jobmanager_job_uptime`	Job uptime	How long the current job has been running without interruption.	ms	Custom
`flink_jobmanager_numRunningJobs`	Number of running jobs	Total jobs currently in a running state on this JobManager.	N/A	Custom
`flink_jobmanager_taskSlotsAvailable`	Available task slots	Slots not currently assigned to any task. A value of 0 means no capacity remains for new tasks.	N/A	Custom
`flink_jobmanager_taskSlotsTotal`	Total task slots	Total slot capacity across all registered TaskManagers.	N/A	Custom
`flink_jobmanager_numRegisteredTaskManagers`	Registered TaskManagers	Number of TaskManagers currently registered with the JobManager. A drop may indicate TaskManager failures.	N/A	Custom

Latency

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_operator_currentEmitEventTimeLag`	Business latency	Time difference between the event timestamp and the wall clock at the point of emission. A high value indicates delays in data pulling or processing.	ms	Custom
`flink_taskmanager_job_task_operator_currentFetchEventTimeLag`	Transmission latency	Time difference between the event timestamp and the wall clock at the point of fetching. A high value indicates delays in data pulling from the source.	ms	Custom
`flink_taskmanager_job_task_operator_watermarkLag`	Watermark latency	Latency at the subtask level, measured as the gap between the current watermark and the wall clock.	ms	Custom
`flink_taskmanager_job_task_currentInputWatermark`	Last received watermark	Timestamp of the most recent watermark received by each task. Useful for detecting event-time skew across subtasks.	N/A	Custom

Throughput -- records

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_numRecordsIn`	Total input records (task)	If this value stops increasing for an extended period, the upstream source may have stopped sending data.	Count	Custom
`flink_taskmanager_job_task_numRecordsOut`	Total output records (task)	If this value stops increasing while input records continue to grow, a logic error in the job code may be dropping data.	Count	Custom
`flink_taskmanager_job_task_operator_numRecordsIn`	Total input records (operator)	If this value stops increasing for a specific operator, the upstream operator may have dropped data.	Count	Custom
`flink_taskmanager_job_task_operator_numRecordsOut`	Total output records (operator)	If this value stops increasing for a specific operator, a logic error may be filtering out all data.	Count	Custom
`flink_taskmanager_job_task_operator_source_numRecordsIn`	Input records (source operator)	Records received by the source operator only. Compare against the total task input to isolate source-level throughput.	Count	Custom
`flink_taskmanager_job_task_operator_sink_numRecordsOut`	Output records (sink operator)	Records written by the sink operator. Compare against operator-level output to verify data reaches the sink.	Count	Custom
`flink_taskmanager_job_task_numRecordsInPerSecond`	Input records per second (task)	Monitor the processing speed of the data stream. Check whether throughput meets expectations under different input loads.	Count/s	Custom
`flink_taskmanager_job_task_numRecordsOutPerSecond`	Output records per second (task)	Monitor the output speed of the data stream. Compare against input rate to detect processing bottlenecks.	Count/s	Custom
`flink_taskmanager_job_task_operator_numRecordsInPerSecond`	Input records per second (operator)	Per-operator input rate. Useful for identifying which operator is the throughput bottleneck.	Count/s	Custom
`flink_taskmanager_job_task_operator_numRecordsOutPerSecond`	Output records per second (operator)	Per-operator output rate. Compare against the input rate of the same operator to measure processing efficiency.	Count/s	Custom
`flink_taskmanager_job_task_operator_source_numRecordsInPerSecond`	Input records per second (source operator)	Generation rate of each data source. If this value drops to 0, the upstream source may have stopped producing data.	Count/s	Custom
`flink_taskmanager_job_task_operator_sink_numRecordsOutPerSecond`	Output records per second (sink operator)	Output rate of each sink. If this value drops to 0, a logic error may be filtering out all data before it reaches the sink.	Count/s	Custom

Throughput -- bytes

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_operator_numBytesIn`	Total input bytes (operator)	Cumulative bytes received from the upstream source. Use this alongside the per-second rate to assess throughput trends.	Bytes	Custom
`flink_taskmanager_job_task_operator_numBytesOut`	Total output bytes (operator)	Cumulative bytes sent to the downstream sink.	Bytes	Custom
`flink_taskmanager_job_task_operator_numBytesInPerSecond`	Input bytes per second (operator)	Byte-level input rate from the upstream source.	Bytes/s	Custom
`flink_taskmanager_job_task_operator_numBytesOutPerSecond`	Output bytes per second (operator)	Byte-level output rate to the downstream sink.	Bytes/s	Custom
`flink_taskmanager_job_task_numBytesInLocalPerSecond`	Input bytes per second (local)	Byte rate for data received from tasks on the same TaskManager.	Bytes/s	Custom
`flink_taskmanager_job_task_numBytesOutPerSecond`	Output bytes per second (task)	Byte-level output rate across all downstream channels.	Bytes/s	Custom
`flink_taskmanager_job_task_numBytesInRemotePerSecond`	Input bytes per second (remote)	Byte rate for data received from tasks on other TaskManagers.	Bytes/s	Custom

Network buffers

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_numBuffersInLocalPerSecond`	Local buffers consumed per second	A high value indicates frequent communication between tasks on the same node.	Count/s	Custom
`flink_taskmanager_job_task_numBuffersInRemotePerSecond`	Remote buffers received per second	Reflects the frequency of cross-TaskManager communication. A high value combined with high latency may indicate network congestion.	Count/s	Custom
`flink_taskmanager_job_task_numBuffersOutPerSecond`	Buffers sent per second	Measures the output pressure of tasks and network bandwidth usage.	Count/s	Custom

Source and sink

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_operator_pendingRecords`	Pending records at source	Number of records in the external system that the source operator has not yet pulled. A growing value indicates the source is falling behind.	Count	Custom
`flink_taskmanager_job_task_operator_sourceIdleTime`	Source idle time	Duration since the source operator last processed a record. A large value indicates a low data generation rate in the external system.	ms	Custom
`flink_taskmanager_job_task_operator_currentSendTime`	Send time for the latest record	Time taken to send the most recent record.	ms	Custom

Checkpoints

Metric	Description	Details	Unit	Type
`flink_jobmanager_job_totalNumberOfCheckpoints`	Total checkpoints	Cumulative number of checkpoints triggered since the job started.	Unit	Custom
`flink_jobmanager_job_numberOfCompletedCheckpoints`	Completed checkpoints	Number of checkpoints that finished successfully.	Unit	Custom
`flink_jobmanager_job_numberOfFailedCheckpoints`	Failed checkpoints	Number of checkpoints that failed. A non-zero value warrants investigation.	unit	Custom
`flink_jobmanager_job_numberOfInProgressCheckpoints`	In-progress checkpoints	Number of checkpoints currently in progress.	Unit	Custom
`flink_jobmanager_job_lastCheckpointDuration`	Last checkpoint duration	Time taken to complete the most recent checkpoint. Possible causes of long durations: large state size, transient network issues, unaligned barriers, or data backpressure.	ms	Custom
`flink_jobmanager_job_lastCheckpointSize`	Last checkpoint size	Size of the most recent checkpoint that was actually uploaded. Compare over time to track state growth.	Bytes	Custom

State access latency

All state access latency metrics report the maximum latency of a single operation, measured in nanoseconds (ns). Use these metrics to identify slow state backend operations that may affect overall job performance.

General

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_state_name_stateClearLatency`	Max latency of a single state clear operation	ns	Custom

Value state

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_state_name_valueStateGetLatency`	Max latency of a single value state read	ns	Custom
`flink_taskmanager_job_task_operator_state_name_valueStateUpdateLatency`	Max latency of a single value state update	nanosecond (ns)	Custom

Aggregating state

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_state_name_aggregatingStateGetLatency`	Max latency of a single aggregating state read	ns	Custom
`flink_taskmanager_job_task_operator_state_name_aggregatingStateAddLatency`	Max latency of a single aggregating state add	ns	Custom
`flink_taskmanager_job_task_operator_state_name_aggregatingStateMergeNamespacesLatency`	Max latency of a single aggregating state merge-namespaces operation	ns	Custom

Reducing state

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_state_name_reducingStateGetLatency`	Max latency of a single reducing state read	ns	Custom
`flink_taskmanager_job_task_operator_state_name_reducingStateAddLatency`	Max latency of a single reducing state add	ns	Custom
`flink_taskmanager_job_task_operator_state_name_reducingStateMergeNamespacesLatency`	Max latency of a single reducing state merge-namespaces operation	ns	Custom

Map state

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_state_name_mapStateGetLatency`	Max latency of a single map state read	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStatePutLatency`	Max latency of a single map state put	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStatePutAllLatency`	Max latency of a single map state putAll	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateRemoveLatency`	Max latency of a single map state remove	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateContainsLatency`	Max latency of a single map state contains check	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateEntriesInitLatency`	Max latency of a single map state entries iterator init	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateKeysInitLatency`	Max latency of a single map state keys iterator init	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateValuesInitLatency`	Max latency of a single map state values iterator init	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateIteratorInitLatency`	Max latency of a single map state iterator init	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateIsEmptyLatency`	Max latency of a single map state isEmpty check	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateIteratorHasNextLatency`	Max latency of a single map state iterator hasNext	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateIteratorNextLatency`	Max latency of a single map state iterator next	ns	Custom
`flink_taskmanager_job_task_operator_state_name_mapStateIteratorRemoveLatency`	Max latency of a single map state iterator remove	ns	Custom

List state

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_state_name_listStateGetLatency`	Max latency of a single list state read	ns	Custom
`flink_taskmanager_job_task_operator_state_name_listStateAddLatency`	Max latency of a single list state add	ns	Custom
`flink_taskmanager_job_task_operator_state_name_listStateAddAllLatency`	Max latency of a single list state addAll	ns	Custom
`flink_taskmanager_job_task_operator_state_name_listStateUpdateLatency`	Max latency of a single list state update	ns	Custom
`flink_taskmanager_job_task_operator_state_name_listStateMergeNamespacesLatency`	Max latency of a single list state merge-namespaces operation	ns	Custom

Sorted map state

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_state_name_sortedMapStateFirstEntryLatency`	Max latency of accessing the first entry of a sorted map state	ns	Custom
`flink_taskmanager_job_task_operator_state_name_sortedMapStateLastEntryLatency`	Max latency of accessing the last entry of a sorted map state	ns	Custom

GeminiDB state storage

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_operator_geminiDB_total_size`	State data size	Total size of state data managed by GeminiDB. Use this metric to locate nodes where state bottlenecks may occur and to verify whether TTL policies are taking effect.	Byte	Custom
`flink_taskmanager_job_task_operator_geminiDB_total_filesize`	State data file size	Size of the state data files on local disk. Monitor this metric to detect disk space pressure caused by large state and to take action before disk runs out.	Byte	Custom

Window metrics

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_operator_numLateRecordsDropped`	Late records dropped	Number of records dropped because they arrived after the window closed.	Unit	Custom
`flink_taskmanager_job_task_operator_lateRecordsDroppedRate`	Late records drop rate	Ratio of late records dropped to total records processed.	N/A	Custom

CDC connector metrics

These metrics apply only to Flink CDC (Change Data Capture) connector jobs. They track progress through the snapshot (full data) and binlog (incremental data) phases.

Phase indicators

Metric	Description	Details	Unit	Type
`flink_jobmanager_job_operator_coordinator_enumerator_isSnapshotting`	Snapshot phase active	Whether the job is in the full-data (snapshot) processing phase.	N/A	Custom
`flink_jobmanager_job_operator_coordinator_enumerator_isBinlogReading`	Binlog phase active	Whether the job is in the incremental-data (binlog) processing phase.	N/A	Custom

Snapshot progress

Metric	Description	Details	Unit	Type
`flink_jobmanager_job_operator_coordinator_enumerator_numTablesRemaining`	Tables remaining (snapshot)	Number of tables not yet processed during the snapshot phase.	Unit	Custom
`flink_jobmanager_job_operator_coordinator_enumerator_numTablesSnapshotted`	Tables pending processing (snapshot)	Number of tables waiting to be processed during the snapshot phase.	Item	Custom
`flink_jobmanager_job_operator_coordinator_enumerator_numSnapshotSplitsProcessed`	Tables processed (snapshot)	Number of tables that have completed snapshot processing.	Item	Custom
`flink_jobmanager_job_operator_coordinator_enumerator_cdcns_schema_table_numSnapshotSplitsProcessed`	Shards processed (snapshot)	Number of shards that have completed snapshot processing.	Unit	Custom
`flink_jobmanager_job_operator_coordinator_enumerator_numSnapshotSplitsRemaining`	Shards remaining (snapshot)	Number of shards waiting to be processed during the snapshot phase.	Unit	Custom
`flink_jobmanager_job_operator_coordinator_enumerator_cdcns_schema_table_numSnapshotSplitsRemaining`	Shards remaining per table (snapshot)	Number of shards waiting to be processed for a specific table.	Unit	Custom

Data volume

Metric	Description	Details	Unit	Type
`flink_taskmanager_job_task_operator_currentReadTimestampMs`	Latest read timestamp	Timestamp of the most recent data record read. Use this to track binlog read progress.	ms	Custom
`flink_taskmanager_job_task_operator_numSnapshotRecords`	Records processed (snapshot)	Total number of records processed during the snapshot phase.	Count	Custom
`flink_taskmanager_job_task_operator_cdcns_schema_table_numRecordsIn`	Records read per table	Total records read from each table across all phases.	Count	Custom
`flink_taskmanager_job_task_operator_cdcns_schema_table_numSnapshotRecords`	Records processed per table (snapshot)	Records processed for each table during the snapshot phase.	Count	Custom

Incremental phase DML and DDL

Metric	Description	Unit	Type
`flink_taskmanager_job_task_operator_numInsertDMLRecords`	INSERT statements processed (total)	Count	Custom
`flink_taskmanager_job_task_operator_numUpdateDMLRecords`	UPDATE statements processed (total)	Items	Custom
`flink_taskmanager_job_task_operator_numDeleteDMLRecords`	DELETE statements processed (total)	Item	Custom
`flink_taskmanager_job_task_operator_numDDLRecords`	DDL statements processed (total)	Item	Custom
`flink_taskmanager_job_task_operator_cdcns_schema_table_numInsertDMLRecords`	INSERT statements processed (per table)	Count	Custom
`flink_taskmanager_job_task_operator_cdcns_schema_table_numUpdateDMLRecords`	UPDATE statements processed (per table)	Items	Custom
`flink_taskmanager_job_task_operator_cdcns_schema_table_numDeleteDMLRecords`	DELETE statements processed (per table)	Count	Custom
`flink_taskmanager_job_task_operator_cdcns_schema_table_numDDLRecords`	DDL statements processed (per table)	Item	Custom

JobManager JVM and system resources

These basic metrics are free of charge when reported from Alibaba Cloud Realtime Compute for Apache Flink. G1 garbage collector metrics are custom metrics.

CPU

Metric	Description	Details	Unit	Type
`flink_jobmanager_Status_JVM_CPU_Load`	CPU load	Recent CPU load of the JobManager JVM. A value consistently above 100% indicates the CPU is saturated, which may cause high response times or system stuttering.	N/A	Basic
`flink_jobmanager_Status_ProcessTree_CPU_Usage`	CPU utilization	CPU time consumed by the JobManager process. 100% = one core fully used, 400% = four cores fully used. If load is high but utilization is low, frequent I/O may be causing processes to enter an uninterruptible sleep state.	N/A	Basic

Memory

Metric	Description	Unit	Type
`flink_jobmanager_Status_JVM_Memory_Heap_Used`	Heap memory used	Bytes	Basic
`flink_jobmanager_Status_JVM_Memory_Heap_Committed`	Heap memory committed	Byte	Basic
`flink_jobmanager_Status_JVM_Memory_Heap_Max`	Heap memory max	Bytes	Basic
`flink_jobmanager_Status_JVM_Memory_NonHeap_Used`	Non-heap memory used	Byte	Basic
`flink_jobmanager_Status_JVM_Memory_NonHeap_Committed`	Non-heap memory committed	Byte	Basic
`flink_jobmanager_Status_JVM_Memory_NonHeap_Max`	Non-heap memory max	Bytes	Basic

Threads

Metric	Description	Details	Unit	Type
`flink_jobmanager_Status_JVM_Threads_Count`	Thread count	Excessive threads consume memory and reduce job stability.	Unit	Basic

Garbage collection

Metric	Description	Details	Unit	Type
`flink_jobmanager_Status_JVM_GarbageCollector_ParNew_Count`	ParNew GC count	Frequent GCs increase memory overhead and degrade job performance. Use this metric for diagnostics and failure troubleshooting.	Count	Basic
`flink_jobmanager_Status_JVM_GarbageCollector_ParNew_Time`	ParNew GC time	Long GC pauses degrade job performance. Use this metric alongside GC count for root-cause analysis.	ms	Basic
`flink_jobmanager_Status_JVM_GarbageCollector_ConcurrentMarkSweep_Count`	CMS GC count	Cumulative number of CMS garbage collections.	Count	Basic
`flink_jobmanager_Status_JVM_GarbageCollector_ConcurrentMarkSweep_Time`	CMS GC time	Cumulative time spent in CMS garbage collection.	ms	Basic
`flink_jobmanager_Status_JVM_GarbageCollector_G1_Young_Generation_Count`	G1 young generation GC count	Cumulative number of G1 young generation garbage collections.	Count	Custom
`flink_jobmanager_Status_JVM_GarbageCollector_G1_Young_Generation_Time`	G1 young generation GC time	Cumulative time spent in G1 young generation garbage collection.	ms	Custom
`flink_jobmanager_Status_JVM_GarbageCollector_G1_Old_Generation_Count`	G1 old generation GC count	Cumulative number of G1 old generation garbage collections.	Count	Custom
`flink_jobmanager_Status_JVM_GarbageCollector_G1_Old_Generation_Time`	G1 old generation GC time	Cumulative time spent in G1 old generation garbage collection.	ms	Custom

Class loading

Metric	Description	Details	Unit	Type
`flink_jobmanager_Status_JVM_ClassLoader_ClassesLoaded`	Classes loaded	Total classes loaded since JVM creation. Excessive class loading increases memory consumption and may degrade performance.	N/A	Basic
`flink_jobmanager_Status_JVM_ClassLoader_ClassesUnloaded`	Classes unloaded	Total classes unloaded since JVM creation. Excessive class unloading increases memory overhead and may degrade performance.	N/A	Basic

TaskManager JVM and system resources

These basic metrics are free of charge when reported from Alibaba Cloud Realtime Compute for Apache Flink. G1 garbage collector metrics are custom metrics.

CPU

Metric	Description	Details	Unit	Type
`flink_taskmanager_Status_JVM_CPU_Load`	CPU load	CPU load of a single TaskManager. In Flink, CPU load is calculated as `CPU Usage / Number of CPU cores`. If `flink_taskmanager_Status_JVM_CPU_Load` exceeds this value, CPU processing may be blocked.	N/A	Basic
`flink_taskmanager_Status_ProcessTree_CPU_Usage`	CPU utilization	CPU time consumed by the TaskManager process. 100% = one core fully used, 400% = four cores fully used. If load is high but utilization is low, frequent I/O may be causing processes to enter an uninterruptible sleep state.	N/A	Basic

Memory

Metric	Description	Unit	Type
`flink_taskmanager_Status_JVM_Memory_Heap_Used`	Heap memory used	Bytes	Basic
`flink_taskmanager_Status_JVM_Memory_Heap_Committed`	Heap memory committed	Byte	Basic
`flink_taskmanager_Status_JVM_Memory_Heap_Max`	Heap memory max	Bytes	Basic
`flink_taskmanager_Status_JVM_Memory_NonHeap_Used`	Non-heap memory used	Byte	Basic
`flink_taskmanager_Status_JVM_Memory_NonHeap_Committed`	Non-heap memory committed	Bytes	Basic
`flink_taskmanager_Status_JVM_Memory_NonHeap_Max`	Non-heap memory max	Bytes	Basic
`flink_taskmanager_Status_ProcessTree_Memory_RSS`	Resident set size (RSS)	Bytes	Basic

Threads

Metric	Description	Details	Unit	Type
`flink_taskmanager_Status_JVM_Threads_Count`	Thread count	Excessive threads consume memory and reduce job stability.	Unit	Basic

Garbage collection

Metric	Description	Details	Unit	Type
`flink_taskmanager_Status_JVM_GarbageCollector_ParNew_Count`	ParNew GC count	Frequent GCs increase memory overhead and degrade job performance. Use this metric for diagnostics and task-level failure troubleshooting.	Count	Basic
`flink_taskmanager_Status_JVM_GarbageCollector_ParNew_Time`	ParNew GC time	Long GC pauses degrade job performance. Use this metric alongside GC count for root-cause analysis.	ms	Basic
`flink_taskmanager_Status_JVM_GarbageCollector_ConcurrentMarkSweep_Count`	CMS GC count	Cumulative number of CMS garbage collections.	Count	Basic
`flink_taskmanager_Status_JVM_GarbageCollector_ConcurrentMarkSweep_Time`	CMS GC time	Cumulative time spent in CMS garbage collection.	ms	Basic
`flink_taskmanager_Status_JVM_GarbageCollector_G1_Young_Generation_Count`	G1 young generation GC count	Cumulative number of G1 young generation garbage collections.	Count	Custom
`flink_taskmanager_Status_JVM_GarbageCollector_G1_Young_Generation_Time`	G1 young generation GC time	Cumulative time spent in G1 young generation garbage collection.	ms	Custom
`flink_taskmanager_Status_JVM_GarbageCollector_G1_Old_Generation_Count`	G1 old generation GC count	Cumulative number of G1 old generation garbage collections.	Count	Custom
`flink_taskmanager_Status_JVM_GarbageCollector_G1_Old_Generation_Time`	G1 old generation GC time	Cumulative time spent in G1 old generation garbage collection.	ms	Custom

Class loading

Metric	Description	Details	Unit	Type
`flink_taskmanager_Status_JVM_ClassLoader_ClassesLoaded`	Classes loaded	Total classes loaded since JVM creation. Excessive class loading increases memory consumption and may degrade performance.	N/A	Basic
`flink_taskmanager_Status_JVM_ClassLoader_ClassesUnloaded`	Classes unloaded	Total classes unloaded since JVM creation. Excessive class unloading increases memory overhead and may degrade performance.	N/A	Basic

Common metric labels

Every Flink metric includes the following Prometheus labels, which you can use to filter and aggregate metrics in PromQL queries:

Label	Description
`vvpNamespace`	Project namespace
`deploymentName`	Deployment name
`deploymentId`	Deployment ID
`jobId`	Job ID

Job health

Latency

Throughput -- records

Throughput -- bytes

Network buffers

Source and sink

Checkpoints

State access latency

General

Value state

Aggregating state

Reducing state

Map state

List state

Sorted map state

GeminiDB state storage

Window metrics

CDC connector metrics

Phase indicators

Snapshot progress

Data volume

Incremental phase DML and DDL

JobManager JVM and system resources

CPU

Memory

Threads

Garbage collection

Class loading

TaskManager JVM and system resources

CPU

Memory

Threads

Garbage collection

Class loading

Common metric labels

See also