PostgreSQL metrics

Database metrics are a comprehensive collection of database workloads and resource usages. The WhaTap monitoring service continuously collects metric data to identify bottlenecks in the database and provides data to help identify the root cause of the performance issue.

For more information about other metrics, see the link.

Indicators	Unit	Description	Category
`active_sessions`	-	Total number of connections for query	Connection
`blk_read_time`	millisecond	Time spent writing data file blocks on the backend	I/O
`blk_write_time`	millisecond	Time spent writing data file blocks on the backend	I/O
`blks_hit`	-	Total count where a disk block was found in the buffer cache and did not require a read	I/O
`blks_hit_ratio`	pct	It indicates the buffer cache hit rate. This can be calculated by sum(`blks_hit`)/sum(`blks_hit` + `blks_read`) * 100.	I/O
`blks_read`	-	Number of disk blocks read from the database	I/O
`buffers_checkpoint`	-	Total number of buffers written by the checkpointer process	Info
`check_point`	-	checkpoint execution count	Troughput
`checkpoint_write_time`	millisecond	Total time in milliseconds for disk write caused by the checkpoint operation	Troughput
`checkpoints_timed`	-	Number of checkpoint operations according to the `checkpoint_timeout` setting	Troughput
`commit_count`	-	Number of transactions in the database committed	Troughput
`conflicts`	-	Number of queries that were aborted because they conflicted with database recovery operations	Troughput
`connect_error`	-	Counted number when the agent failed to access the database	Connection
`deadlock_count`	-	Total number of deadlocks detected in the database	Performance
`dml_count`	-	Number of DML executions	Troughput
`index_hit_ratio`	pct	Rate at which indexes were read from the buffer cache	Performance
`lock_wait_sessions`	-	Number of sessions waiting for a lock	Performance
`long_running_sessions`	-	Number of sessions running for more than the long running time (default:10 seconds)	Performance
`max_connections`	-	This value is set in `pg_settings` for the maximum number of possible connections. If it is not enough, `FATAL ERROR: too many clients already.` Errors may occur.	Info
`node_state`	-	This database consists of replication relationships. `1`: Master, `2`: Slave.	Info
`numbackends`	-	Number of connected backends	Info
`replication_count`	-	Number of replication databases (Slave)	Performance
`replication_delay`	-	Delay time reflected in the replication database for the replication structure, this is a metric measured in the slave DB	Performance
`rollback_count`	-	Total number of transactions in the database rolled back	Performance
`table_hit_ratio`	-	Rate at which the tables were read from the buffer cache	Performance
`temp_bytes`	-	Total amount of data written to temporary files by queries in the database	Performance
`temp_files`	-	Total number of temporary files generated by queries in the database	Performance
`total_runtime`	-	Sum of runtimes (elapsed time) for all active sessions	Performance
`total_sessions`	-	Total number of connected sessions	Connection
`tup_deleted`	-	Total number of rows deleted by the queries in the database	Troughput
`tup_fetched`	-	Total number of rows fetched by the queries in the database	Troughput
`tup_inserted`	-	Total number of rows added by the queries in the database	Troughput
`tup_returned`	-	Number of rows returned by the queries in the database	Troughput
`tup_updated`	-	Total number of rows updated by the queries in the database	Troughput
`vacuum_running_count`	-	Number of processes for vacuuming	Performance
`wait_sessions`	-	Total number of queued connections	Performance
`seq_scan`	-	Method to read records during full table scanning. Select this when any index does not exist or if an index exists, the range to be read is wide. This metric has the values queried every minute.	Performance
`idx_scan`	-	With this method, table records are accessed using keys stored in index leaf blocks. This metric has values retrieved every minute.	Performance
`wal`	-	Number of WAL files. Spikes in the WAL file counts are usually caused by the large amount of changes in a short period or vacuum-like maintenance operations that generate temporary tables and objects. They should slowly return to normal levels. This usually results in lots of disk I/O and CPU activity, slowing the application query's execution until things get back to normal. This metric has the values queried every minute.	Performance
`cpu(xos2)`	pct	CPU utilization, `cpu_user` + `cpu_sys` + `cpu_iowait`.	XOS
`cpu_user(xos2)`	pct	Time spent for CPU to execute user code	XOS
`cpu_sys(xos2)`	pct	Time spent for CPU to execute the kernel code	XOS
`cpu_iowait(xos2)`	pct	CPU time waited for IO	XOS
`cpu_idle(xos2)`	pct	CPU processor's idle time	XOS
`mem(xos)`	pct	Memory utilization	XOS
`mem_total(xos)`	MB	Total amount of memory	XOS
`mem_free(xos)`	MB	Amount of memory not in use by the system	XOS
`mem_used(xos)`	MB	Amount of memory in use by the system	XOS
`mem_buffcache(xos)`	MB	Amount of memory used by the buff/cache	XOS
`mem_swaptotal(xos)`	MB	Set this in preparation for the case where the total Swap size and physical memory are insufficient.	XOS
`mem_swapfree(xos)`	MB	Because the unused Swap area and Swap area are secured in disk space, the system performance may drop drastically while using Swap.	XOS
`mem_available(xos)`	MB	Available memory area	XOS
`disk_reads(xos)`	block	Amount of data read from disk	XOS
`disk_read_bytes(xos)`	byte	Amount of data read from disk	XOS
`disk_writes(xos)`	block	Amount of data written to disk	XOS
`disk_write_bytes(xos)`	byte	Amount of data written to disk	XOS
`net_recv_bytes(xos)`	byte	Amount of data received over the network	XOS
`net_recv_packets(xos)`	-	Number of data packets received over the network	XOS
`net_send_bytes(xos)`	byte	Amount of data transmitted over the network	XOS
`net_send_packets(xos)`	-	Number of data packets transmitted over the network	XOS