Monitoring Feature

Last updated: 2018-09-14 12:13:46

PDF

Performance Monitoring

PostgreSQL provides various performance monitoring metrics to make it easy for users to view and obtain the operation information of instances. You can check the information in Instance Management -> System Monitoring in the PostgreSQL console.

pg_stat_statements

You can view detailed performance metrics of pg from the pg_stat_statements view.

Available Monitoring Metrics

Metric Name Metric API Description Unit
CPU utilization cpu CPU utilization of the instance. This may go beyond 100% due to flexible CPU limitation strategies when idle %
Occupied storage storage Disk storage space occupied by the instance GB
Disk IOPS iops IOPS (number of requests per second) of the instance requests per second
Input traffic in_flow Input traffic generated by data read/write operations for the instance KB per second
Output traffic out_flow Output traffic generated by data read/write operations for the instance KB per second
Number of connections connections Historical trend for the number of active connections of the instance -
Number of requests read_write_calls Total number of read/write requests (addition, deletion, modification and query) per minute requests per minute
Number of read requests read_calls Total number of read requests per minute requests per minute
Number of write requests write_calls Total number of write requests per minute requests per minute
Number of other requests other_calls Total number of requests other than read/write requests (such as Drop), accumulated by the minute requests per minute
Cache hit rate hit_percent Data cache hit rate %
Average execution latency sql_runtime_avg The average execution time of all SQL requests (SQLs in transactions not included). ms
TOP 10 longest execution latencies sql_runtime_avg Average execution latency of top 10 longest SQL requests ms
TOP 10 shortest execution latencies sql_runtime_min Average execution latency of top 10 shortest SQL requests ms
Number of remaining XID remain_xid Number of remaining Transaction IDs. There are a maximum of 2^32 Transaction Ids. It is recommended to execute "vacuum full" manually if the number falls below 1,000,000. -
Synchronization difference between master/slave XLOG xlog_diff (Sample is taken every minute) The synchronization difference between the master XLOG and the slave XLOG. This represents synchronization delay, and lower is better. byte