Amazon Aurora MySQL Engine Metrics Dashboard - Track Queries

Use this dashboard to monitor what the MySQL engine reports about itself on an Aurora cluster: connections and threads, query workload, the InnoDB buffer pool, rows and locks, and InnoDB I/O, logs, and temporary resources.

The OpenTelemetry Collector mysql receiver connects to the database and reads SHOW GLOBAL STATUS and the InnoDB tables directly. This is the query-level view that CloudWatch cannot provide.

Dashboard Coverage

Use these panels to:

• Size the thread cache: Created climbing while connected stays flat means the thread cache is too small.
• Find queries without an index: A large read_rnd_next share on Handlers, or full joins, means scans rather than seeks.
• Judge buffer pool fit: Watch free pages trend toward zero and read operations rise as the working set outgrows memory.
• Diagnose lock contention: Read row lock waits and accumulated wait time against table locks that had to wait.
• Catch spills to disk: Rising disk_tables on Temporary Resources means queries are spilling out of memory.
• Spot restarts and failovers: A drop on Uptime marks the instance restarting.

Metrics Included

Connections & Threads

• Threads: Threads cached, connected, created, and running (mysql.threads).
• Connection Attempts: Connection attempts per interval (mysql.connection.count).
• Connection Errors: Connection errors per interval, by cause (mysql.connection.errors).
• Client Network I/O: Bytes sent to and received from clients (mysql.client.network.io).

Query Workload

• Commands: Statements executed per interval by command type (mysql.commands).
• Queries: Statements executed per interval (mysql.query.count).
• Slow Queries: Statements exceeding long_query_time (mysql.query.slow.count).
• Prepared Statements: Prepared-statement commands per interval (mysql.prepared_statements).
• Handlers: Storage-engine handler calls per interval (mysql.handlers).
• Joins: Join plans per interval by type (mysql.joins). full and full_range are joins running without a usable index.
• Sorts: Sort work per interval (mysql.sorts). Rising merge_passes means sort_buffer_size is too small.

InnoDB Buffer Pool

• Buffer Pool Usage: Bytes held as clean and dirty pages (mysql.buffer_pool.usage).
• Buffer Pool Limit: Configured innodb_buffer_pool_size (mysql.buffer_pool.limit). On Aurora the instance class or ACU range sets this.
• Buffer Pool Pages: Pages by kind (mysql.buffer_pool.pages).
• Buffer Pool Data Pages: Clean and dirty data pages (mysql.buffer_pool.data_pages).
• Buffer Pool Operations: Operations per interval (mysql.buffer_pool.operations). reads are requests that missed the pool and went to storage.
• Buffer Pool Page Flushes: Pages flushed per interval (mysql.buffer_pool.page_flushes).

Rows & Locks

• Row Operations: Rows read, inserted, updated, and deleted per interval (mysql.row_operations).
• Row Locks: Row-lock waits and accumulated wait time (mysql.row_locks).
• Table Locks: Locks granted immediately against those that had to wait (mysql.locks).
• Page Operations: InnoDB pages created, read, and written per interval (mysql.page_operations).

InnoDB I/O, Logs & Resources

• InnoDB I/O Operations: Reads, writes, and fsyncs per interval (mysql.operations).
• Redo Log Operations: Redo log writes, write requests, and waits (mysql.log_operations). Waits above zero mean the log buffer is too small.
• Doublewrite Buffer: Doublewrite pages and write operations per interval (mysql.double_writes).
• Opened Resources: Files, tables, and table definitions opened per interval (mysql.opened_resources). Sustained table opens mean table_open_cache is too small.
• Temporary Resources: Temporary tables and files created per interval (mysql.tmp_resources).
• Uptime: Time since the server started (mysql.uptime).

Dashboard Variables

Use these filter variables:

• mysql_instance_endpoint: Filter by the endpoint each receiver scrapes (mysql.instance.endpoint). Point one receiver at each instance endpoint so every instance is its own series.

Panels That Read Empty

Some panels stay empty until the matching condition exists, which is expected rather than a misconfiguration:

• Connection Attempts, Connection Errors, Client Network I/O, Commands, Queries, Slow Queries, and Joins are opt-in metrics. Enable them on the receiver as shown in the setup guide.
• Slow Queries also needs slow_query_log enabled and a long_query_time your workload exceeds.
• Connection Errors and Row Locks read zero on a healthy cluster.

Limitations

Aurora MySQL has no binary log replication between cluster instances, so SHOW REPLICA STATUS returns no rows and the receiver's mysql.replica.time_behind_source and mysql.replica.sql_delay metrics never populate. No panel here uses them. Use Aurora Replica Lag on the CloudWatch dashboard instead.

Statement-event metrics need performance_schema, which Aurora MySQL sets to OFF by default. Turning it on requires a DB parameter group and an instance reboot, so this dashboard does not depend on it.

Next Steps

• Add the CloudWatch dashboard for cluster capacity, replica lag, and volume growth.
• Trace slow queries back to the application code that ran them.
• Set up alerts on slow query rate, row lock waits, and buffer pool pressure.

Related Dashboards

• Amazon Aurora MySQL CloudWatch
• Amazon Aurora PostgreSQL Engine Metrics
• MySQL