Performance metrics

PolarDB exposes performance metrics for each instance node and the PolarProxy layer. Use these metrics to configure monitoring alerts, query time-series data through the API, or diagnose performance bottlenecks.

Three metric sets are available, each described in a table below:

PolarDB for MySQL — engine and storage metrics for MySQL-compatible clusters
PolarDB for PostgreSQL — engine and storage metrics for PostgreSQL-compatible clusters
PolarProxy — proxy-layer metrics, available only when a cluster endpoint (database proxy) is enabled

Each metric is identified by its Metric name, which is the field name used in API calls and monitoring queries. All metrics use the data type double.

Performance metrics for PolarDB for MySQL

The following metrics cover resource utilization, InnoDB engine internals, DML throughput, I/O, and replication health for PolarDB for MySQL instances.

Performance metric	Metric	Data type	Unit	Description
PolarDBDiskUsage	mean_data_size	double	MB	Storage space used by data.
	mean_log_size	double	MB	Storage space used by logs.
	mean_sys_dir_size	double	MB	Storage space used by system files.
	mean_tmp_dir_size	double	MB	Storage space used by temporary files.
	mean_redolog_size	double	MB	Storage space used by redo logs.
	mean_binlog_size	double	MB	Storage space used by binary logs.
	mean_other_log_size	double	MB	Storage space used by other log files.
	mean_undolog_size	double	MB	Storage space used by undo logs.
PolarDBQPSTPS	mean_qps	double	Count	Number of requests per second.
	mean_tps	double	Count	Number of transactions per second.
	mean_mps	double	Count	Number of operations per second.
PolarDBCPU	cpu_ratio	double	Percent	CPU utilization (%).
PolarDBMemory	mem_ratio	double	Percent	Memory usage (%).
PolarDBMemoryForPCU	pcu_mem_usage	double	Percent	Memory usage relative to the allocated PolarDB Capacity Units (PCUs).
PolarDBCPUForPCU	pcu_cpu_usage	double	Percent	CPU utilization relative to the allocated PCUs.
PolarDBPCU	pcu	double	Cores	Number of PCU cores currently in use.
PolarDBConnections	mean_active_session	double	Count	Number of active connections.
	thread_pool_running_threads	double	Item	Number of threads currently running in the thread pool.
	mean_total_session	double	Item	Total number of current connections.
	tp_thread_count	double	Count	Total number of threads in the thread pool.
PolarDBNetworkTraffic	mean_input_traffic	double	KB/s	Inbound network traffic per second.
PolarDBNetworkTraffic	mean_output_traffic	double	KB/s	Outbound network traffic per second.
PolarDBInnoDBBufferRatio	mean_innodb_buffer_dirty_ratio	double	Percent	Dirty page ratio of the InnoDB buffer pool (%). A high value indicates memory pressure or insufficient checkpoint frequency.
	mean_innodb_buffer_read_hit	double	Percent	Read hit rate of the InnoDB buffer pool (%). A low value indicates frequent disk reads.
	mean_innodb_buffer_use_ratio	double	Percent	Utilization of the InnoDB buffer pool (%).
PolarDBInnoDBDataReadWrite	mean_innodb_data_read	double	Byte	Data read from the InnoDB storage engine per second.
PolarDBInnoDBDataReadWrite	mean_innodb_data_written	double	Byte	Data written to the InnoDB storage engine per second.
PolarDBInnoDBBufferRequests	mean_innodb_buffer_pool_read_requests	double	Count	Read requests served by the buffer pool per second.
PolarDBInnoDBBufferRequests	mean_innodb_buffer_pool_write_requests	double	Count	Write requests to the buffer pool per second.
PolarDBInnoDBLogWrites	mean_innodb_log_write_requests	double	Count	Physical InnoDB log writes per second.
PolarDBInnoDBLogWrites	mean_innodb_os_log_fsyncs	double	Times	InnoDB log fsync calls per second.
PolarDBCreatedTempDiskTable	mean_created_tmp_disk_tables	double	Unit	Temporary disk tables created per second. A high value may indicate missing indexes or insufficient `tmp_table_size`.
PolarDBCOMDML	mean_com_delete	double	Item	DELETE statements executed per second.
	mean_com_delete_multi	double	Item	Multi-table DELETE statements executed per second.
	mean_com_insert	double	Unit	INSERT statements executed per second.
	mean_com_insert_select	double	Unit	INSERT...SELECT statements executed per second.
	mean_com_replace	double	Statements/second	REPLACE statements executed per second.
	mean_com_replace_select	double	Unit	REPLACE...SELECT statements executed per second.
	mean_com_select	double	Statements/second	SELECT statements executed per second.
	mean_com_update	double	Statements/second	UPDATE statements executed per second.
	mean_com_update_multi	double	Unit	Multi-table UPDATE statements executed per second.
PolarDBRowDML	mean_innodb_rows_deleted	double	Row	Rows deleted per second by InnoDB.
	mean_innodb_rows_inserted	double	Row	Rows inserted per second by InnoDB.
	mean_innodb_rows_read	double	Row	Rows read per second by InnoDB.
	mean_innodb_rows_updated	double	Row	Rows updated per second by InnoDB.
PolarDBIOSTAT	mean_iops_r	double	IOPS	Read IOPS.
	mean_iops_w	double	IOPS	Write IOPS.
	mean_iops	double	IOPS	Total IOPS.
	mean_io_throughput_r	double	MB	Read I/O throughput.
	mean_io_throughput_w	double	MB	Write I/O throughput.
	mean_io_throughput	double	MB	Total I/O throughput.
PolarDBReplicaLag	mean_replication_delay	double	Seconds	Replication delay between the primary node and secondary nodes.

Performance metrics for PolarDB for PostgreSQL

The following metrics cover resource utilization, WAL and replication health, transaction state, buffer cache, temporary file usage, and DML throughput for PolarDB for PostgreSQL instances.

Performance metric	Metric	Data type	Unit	Description
PolarDBCPU	cpu_ratio	double	Percent	Total CPU utilization.
	cpu_sys_ratio	double	Percent	CPU utilization in kernel mode.
	cpu_user_ratio	double	Percent	CPU utilization in user mode.
PolarDBMemory	mem_ratio	double	Percent	Memory usage.
PolarDBMemoryForPCU	pcu_mem_usage	double	Percent	Memory usage relative to the allocated PCUs.
PolarDBCPUForPCU	pcu_cpu_usage	double	Percent	CPU utilization relative to the allocated PCUs.
PolarDBPCU	pcu_by_cpu_cores	double	Cores	Number of PCU cores currently in use.
PolarDBIOSTAT	mean_io_throughput_w	double	MB	Write throughput.
	mean_io_throughput_r	double	MB	Read throughput.
	mean_io_throughput	double	MB	Total throughput.
	mean_iops_w	double	Count	Write IOPS.
	mean_iops_r	double	Count	Read IOPS.
	mean_iops	double	Time	Total IOPS.
PolarDBDiskUsage	mean_tmp_size	double	MB	Temporary space usage.
	mean_log_size	double	MB	Local log usage.
	mean_wal_size	double	MB	Space used by Write-Ahead Logging (WAL) logs.
	mean_ins_total_size	double	MB	Total space used by the cluster.
	mean_data_size	double	MB	Total space used by user data.
PolarDBReplication	mean_max_slot_wal_delay_in_mb	double	MB	Maximum replication slot delay. Monitor this metric to detect stalled consumers that prevent WAL segment recycling.
	mean_logical_rep_latency_in_mb	double	MB	Logical replication delay.
	mean_send_latency_in_mb	double	MB	WAL send latency on the primary node.
	mean_replay_latency_in_mb	double	MB	WAL replay latency on the secondary node.
PolarDBSQL	mean_five_seconds_two_pc_transactions	double	Unit	Number of two-phase commit transactions running for 5 seconds.
	mean_three_seconds_two_pc_transactions	double	Unit	Number of two-phase commit transactions running for 3 seconds.
	mean_one_second_two_pc_transactions	double	Item	Number of two-phase commit transactions running for 1 second.
	mean_five_seconds_long_transactions	double	Unit	Number of transactions running for 5 seconds.
	mean_three_seconds_transactions	double	Unit	Number of transactions running for 3 seconds.
	mean_one_second_transactions	double	Unit	Number of transactions running for 1 second.
	mean_five_seconds_idle_transactions	double	Unit	Number of transactions idle for 5 seconds.
	mean_three_seconds_idle_transactions	double	Item	Number of transactions idle for 3 seconds.
	mean_one_second_idle_transactions	double	Unit	Number of transactions idle for 1 second.
	mean_five_seconds_executing_sqls	double	Item	Number of SQL statements running for 5 seconds.
	mean_three_seconds_executing_sqls	double	Item	Number of SQL statements running for 3 seconds.
	mean_one_second_executing_sqls	double	Item	Number of SQL statements running for 1 second.
PolarDBSlowSQLTime	max_executing_sql_last_time	double	ms	Running time of the slowest active SQL statement.
PolarDBReplicationSlots	total_slots	double	Unit	Total number of replication slots.
PolarDBReplicationSlots	inactive_slots	double	Unit	Number of inactive replication slots. Inactive slots that hold back WAL can cause disk space issues.
PolarDBVacuum	mean_db_age	double	xids	Maximum age of the database in transaction IDs (xids). Monitor this metric to assess transaction ID wraparound risk.
PolarDBTransaction	mean_long_two_pc_transactions	double	Count	Number of long-running two-phase commit transactions.
	mean_two_pc_transactions	double	Unit	Total number of two-phase commit transactions.
	mean_long_transactions	double	Unit	Number of long-running transactions.
	mean_long_idle_transactions	double	Unit	Number of long-running idle transactions.
	mean_idle_transactions	double	Count	Number of idle transactions.
	mean_waiting_transactions	double	Unit	Number of transactions waiting for a lock.
	mean_active_transactions	double	Unit	Number of active transactions.
PolarDBBuffer	mean_blks_hit_delta	double	Count	Block cache hits per second.
	mean_blks_read_delta	double	Blocks/second	Blocks read from disk per second. A high value relative to `mean_blks_hit_delta` indicates a low cache hit rate.
	mean_buffers_alloc	double	Count	Total number of cache buffers allocated.
	mean_buffers_backend_fsync	double	Count	Total number of fsync calls performed by backends.
	mean_buffers_backend	double	Count	Total number of cache buffers written directly by backends.
	mean_buffers_clean	double	Count	Total number of cache buffers written by the checkpointer process.
	mean_hit_ratio	double	Percent	Block cache hit rate (%).
PolarDBTemp	mean_temp_bytes	double	Bytes	Data written to temporary files per second. A high value indicates that queries are spilling to disk.
PolarDBTemp	mean_temp_files	double	Item	Temporary files created per second.
PolarDBQPSTPS	mean_deadlocks_delta	double	Number of times	Deadlocks detected per second.
	mean_rollbacks_delta	double	Count	Rollbacks per second.
	mean_commits_delta	double	Count	Commits per second.
	mean_tps	double	Times	Transactions processed per second.
PolarDBRowDML	mean_tup_deleted_delta	double	Row	Rows deleted per second.
	mean_tup_updated_delta	double	Row	Rows updated per second.
	mean_tup_inserted_delta	double	Row	Rows inserted per second.
	mean_tup_fetched_delta	double	Row	Rows fetched per second (rows returned by index scans).
	mean_tup_returned_delta	double	Row	Rows returned per second (rows scanned by sequential scans).
PolarDBConnections	mean_idle_connection	double	Unit	Number of idle connections.
	mean_waiting_connection	double	Unit	Number of connections waiting for a lock.
	mean_active_session	double	Unit	Number of active sessions.
	mean_total_session	double	Unit	Total number of sessions.

Performance metrics for PolarProxy

PolarProxy is the proxy layer that handles connection pooling and read/write splitting for cluster endpoints. The following metrics are available only when a cluster endpoint is enabled for your PolarDB cluster.

Performance metric	Metric	Data type	Unit	Description
PolarProxy_CpuUsage	docker_container_cpu	double	Percent	CPU utilization of the PolarProxy container.
PolarProxy_Cps	service_connections_ps	double	Connections/second	New connections to PolarProxy per second.
PolarProxy_CurrentConns	service_current_connections	double	Unit	Current connections to PolarProxy.
PolarProxy_UpFlows	service_upstream_netflow	double	KB/s	Inbound traffic to PolarProxy per second.
PolarProxy_DownFlows	service_downstream_netflow	double	KB/s	Outbound traffic from PolarProxy per second.
PolarProxy_Qps	service_queries	double	Requests/second	Requests received by PolarProxy per second.
PolarProxy_DBConns	server_connections	double	Unit	Connections from PolarProxy to each DB node.
PolarProxy_DBQps	server_queries	double	Requests/second	Request rate from PolarProxy to each DB node.
PolarProxy_DBActionOps	server_active_operations	double	Item	Active requests from PolarProxy to each DB node.
PolarProxy_LsnNotMatch	service_queries_lsn_not_match	double	Requests/second	Rate at which requests are routed to the primary node because the consistency level is not met for a cluster endpoint. A sustained high value may indicate that read replicas are lagging.
PolarProxy_QueriesInTrx	service_queries_in_trx	double	Requests/second	Rate at which requests are routed to the primary database because they are part of a transaction for a cluster endpoint.
PolarProxy_Rt	service_proxy_rt	double	Microseconds	Round-trip time (RT) at the PolarProxy layer.
PolarProxy_Db_Rt	service_db_rt	double	Microseconds	Average time for a DB node to process a request.
PolarProxy_Full_Rt	service_full_rt	double	Microseconds	Average total time to process a request end-to-end.