How to view performance monitoring metrics for PolarDB - PolarDB

The PolarDB console provides performance metrics with one-second granularity. These metrics help you monitor cluster status and use detailed data to quickly resolve operational issues.

Procedure

Log on to the PolarDB console.
In the navigation pane on the left, click Clusters.
In the upper-left corner, select the region where the cluster is deployed.
Find the target cluster and click its ID.
In the left navigation bar, select Performance Monitoring.
You can view monitoring information for Cluster, Database Agent, Compute Node, Advanced Monitoring, or Performance Events based on your business requirements. For more information, see Metric descriptions.
- To view cluster performance monitoring data, click the Cluster tab and set a time range at the top of the page.
- Database Proxy performance monitoring: Click the Database Agent tab, set a time range and select the corresponding cluster endpoint at the top of the page to view the monitoring data.
- To view compute node performance monitoring data, click the Compute Node tab, select a node, and set a time range at the top of the page.
- To view advanced monitoring data, click the Advanced Monitoring tab, select a node, and set a time range at the bottom of the page.
- To view performance events, click the Performance Events tab, select a node, and set a time range at the bottom of the page.

Note

The Real-time Monitoring Dashboard feature of Database Autonomy Service (DAS) now supports PolarDB for MySQL. This feature helps you quickly identify abnormal clusters for further optimization and analysis.

Metrics

Category	Metric	Description
Cluster	Storage capacity	The size of local binary logs, redo logs, and other log files, as well as the usage of data space, system space, and temporary space. Contact Alibaba Cloud technical support for this metric.
	QPS	The queries per second (QPS) for each node. This metric is calculated from the `Questions` value in the output of the `show global status` command.
	TPS	The transactions per second (TPS) for each node. This metric is calculated from the `Com_commit` and `Com_rollback` values in the output of the `show global status` command.
	MPS	The data manipulations per second (MPS) for each node. This metric is calculated from the `Com_insert`, `Com_insert_select`, `Com_update`, `Com_update_multi`, `Com_delete`, `Com_delete_multi`, `Com_replace`, and `Com_replace_select` values in the output of the `show global status` command.
	CPU utilization	The CPU utilization of each node. Contact Alibaba Cloud technical support for this metric.
	memory usage	The memory usage of each node. Contact Alibaba Cloud technical support for this metric.
	Current connections	The number of active connections on each node. This metric is calculated from the statistics in the output of the `show processlist` command.
	Redo log writes per second	The amount of redo logs written per second on the primary node.
Compute node	QPS	The queries per second (QPS) for the selected node. This metric is calculated from the `Questions` value in the output of the `show global status` command.
	TPS	The transactions per second (TPS) for the selected node. This metric is calculated from the `Com_commit` and `Com_rollback` values in the output of the `show global status` command.
	MPS	The data manipulations per second (MPS) for the selected node. This metric is calculated from the `Com_insert`, `Com_insert_select`, `Com_update`, `Com_update_multi`, `Com_delete`, `Com_delete_multi`, `Com_replace`, and `Com_replace_select` values in the output of the `show global status` command.
	CPU utilization	The CPU utilization of the selected node. Contact Alibaba Cloud technical support for this metric.
	memory usage	The memory usage of the selected node. Contact Alibaba Cloud technical support for this metric.
	Current connections	The total number of current connections and active connections on the selected node. This metric is calculated from the statistics in the output of the `show processlist` command.
	Operations	The number of operations per second on the selected node, such as DELETE, INSERT, UPDATE, and REPLACE. This metric is calculated from corresponding values prefixed with `Com_` in the output of the `show global status` command.
	InnoDB buffer pool	The dirty page ratio, read hit ratio, and utilization of the buffer pool on the selected node. This metric is calculated from values such as `Innodb_buffer_pool_pages_dirty`, `Innodb_buffer_pool_pages_total`, `Innodb_buffer_pool_reads`, and `Innodb_buffer_pool_read_requests` in the output of the `show global status` command.
	I/O throughput	The total, read, and write I/O throughput of the selected node. Contact Alibaba Cloud technical support for this metric.
	IOPS	The I/O operations per second (IOPS) of the selected node, including total, read, and write IOPS. Contact Alibaba Cloud technical support for this metric.
	Network	The inbound and outbound traffic per second for the selected node. This metric is calculated from the `Bytes_received` and `Bytes_sent` values in the output of the `show global status` command.
	Rows scanned	The number of rows inserted, read, updated, and deleted per second on the selected node. This metric is calculated from the `Innodb_rows_deleted`, `Innodb_rows_inserted`, `Innodb_rows_read`, and `Innodb_rows_updated` values in the output of the `show global status` command.
	InnoDB data reads/writes	The amount of data read from and written to the storage engine per second on the selected node. This metric is calculated from the `Innodb_data_read` and `Innodb_data_written` values in the output of the `show global status` command.
	InnoDB buffer pool requests	The number of read and write requests per second for the buffer pool of the selected node. This metric is calculated from the `Innodb_buffer_pool_read_requests` and `Innodb_buffer_pool_write_requests` values in the output of the `show global status` command.
	InnoDB log writes	The number of log write requests per second and the number of times data is flushed to disk per second on the selected node. This metric is calculated from the `Innodb_log_write_requests` and `Innodb_os_log_fsyncs` values in the output of the `show global status` command.
	Temporary tables	The number of temporary tables created per second on the selected node. This metric is calculated from the `Created_tmp_disk_tables` value in the output of the `show global status` command.
PolarProxy	CPU utilization	The CPU utilization of PolarProxy.
	Current connections	The number of active connections on PolarProxy.
	Inbound traffic per second	The inbound traffic per second on PolarProxy.
	Outbound traffic per second	The outbound traffic per second on PolarProxy.
	Requests per second	The number of requests per second on PolarProxy.
Advanced monitoring	PolarDB CPU utilization/memory usage	The CPU utilization and memory usage of the selected node.
	TPS/QPS	The transactions per second (TPS) and queries per second (QPS) of the selected node.
	Session	The total number of current sessions and active sessions on the selected node.
	Traffic throughput	The average number of bytes sent to and received from all clients per second on the selected node.
	Temporary tables	The number of temporary tables created per second on the selected node.
	Executions	The average number of executions per second for statements such as replace, multi_delete, multi_update, insert_select, delete, select, insert, replace_select, and update on the selected node.
	Replication latency	The latency of data replication from the primary node to the selected read-only node.
	InnoDB data I/O throughput	The average number of bytes read and written by InnoDB per second on the selected node.
	InnoDB buffer pool requests	The average number of page reads (logical reads) from and page writes to the buffer pool by InnoDB per second on the selected node.
	InnoDB buffer pool hit ratios	The InnoDB buffer pool usage, dirty page ratio, and read hit ratio of the selected node.
	InnoDB redo log writes	The average number of physical log writes and fsyncs per second on the selected node.
	InnoDB row operations	The average number of rows read, updated, inserted, and deleted by InnoDB per second on the selected node.
	PolarDB storage usage	The storage space used by data such as redo logs, binary logs, system data, business data, log files, and temporary files on the selected node.
	PolarDB IOPS	The I/O operations per second (IOPS) of the selected node, including total IOPS, read IOPS, write IOPS, total I/O throughput, read I/O throughput, and write I/O throughput.

FAQ

Q: Why does the QPS metric show approximately 10 even when there is no traffic to my cluster?

A: Background system tasks, including monitoring, log collection, and administrative operations, generate about 10 queries per second. This activity has a negligible impact on cluster performance.
Q: What should I do if CPU utilization is too high?
A: Investigate the issue by following these steps:
- Check for a large number of slow SQL queries. We recommend that you optimize any slow SQL queries and then test again. To learn how to view and optimize slow SQL queries, see Slow SQL queries.
- Check if the CPU utilization curve correlates with the QPS or TPS curves. If so, the issue is likely caused by high transaction concurrency. In this case, consider upgrading your cluster specifications. For more information, see manual scaling.
- If there are no slow SQL queries and the CPU utilization curve does not follow the QPS or TPS curves, contact Alibaba Cloud technical support for assistance.
Q: What should I do if the number of current connections is much higher than the number of active connections?

A: You can try setting lower values for the wait_timeout and interactive_timeout parameters to release idle connections more quickly. However, the best practice is to ensure your application closes connections promptly after use.

Related APIs

API	Description
DescribeDBClusterPerformance	Queries the performance data of a PolarDB cluster.
DescribeDBNodePerformance	Queries the performance data of a node in a PolarDB cluster.
DescribeDBClusterMonitor	Queries the collection frequency of monitoring data for a PolarDB cluster.