Data of various types is collected inside AnalyticDB for MySQL. The collected data is used by the monitoring and alerting system to trace and troubleshoot issues.
Time series data collection
- Time series data in processes
- Computing resources: CPU utilization.
- Storage resources: I/O throughput, I/O usage, IOPS, used storage, and storage usage.
- Memory resources: total memory, available memory, and memory usage.
- Time series data in services
Time series data of various types are collected in services based on different roles of processes.
- Time series data related to controllers: time series metrics related to queries and data writes. Time series metrics related to queries include queries per second (QPS), query response time, queuing time, planning time, and query failure rates. Time series metrics related to data writes include transactions per second (TPS), number of BUILD INDEX tasks, and write response time.
- Time series data related to executors: number of running splits, memory pool in use, and memory pool usage.
- Time series data related to workers: shard status.
Log data collection
All kinds of internal logs are collected and sent to Log Service for centralized storage and processing. The following log data is included:
- Audit logs: record all operations, such as DDL and DML operations.
- Query details logs: record detailed query information during execution, including the execution plan and memory and CPU consumption on each node.
- Storage logs: record detailed information generated by a query in the process of reading data from the storage layer, such as the time consumed by retrieving indexes and reading blocks.