Amazon Relational Database Service
User Guide (API Version 2014-10-31)

Performance Insights Counters

With counter metrics, you can customize the Performance Insights dashboard to include up to 10 additional graphs. These graphs that show a selection of dozens of operating system and database performance metrics. This information can be correlated with database load to help identify and analyze performance problems.

Performance Insights Operating System Counters

The following operating system counters are available with Performance Insights for Aurora PostgreSQL. You can find definitions for these metrics in Viewing Enhanced Monitoring by Using CloudWatch Logs.

Counter Type

active

memory

buffers memory
cached memory
dirty memory
free memory
inactive memory
hugePagesFree memory
hugePagesRsvd memory
hugePagesSize memory
hugePagesSurp memory
hugePagesTotal memory
mapped memory
pageTables memory
slab memory
total memory
writeback memory
guest cpuUtilization
idle cpuUtilization
irq cpuUtilization
nice cpuUtilization
steal cpuUtilization
system cpuUtilization
total cpuUtilization
user cpuUtilization
wait cpuUtilization
avgQueueLen diskIO
avgReqSz diskIO
await diskIO
readIOsPS diskIO
readKb diskIO
readKbPS diskIO
rrqmPS diskIO
tps diskIO
util diskIO
writeIOsPS diskIO
writeKb diskIO
writeKbPS diskIO
wrqmPS diskIO
blocked tasks
running tasks
sleeping tasks
stopped tasks
total tasks
zombie tasks
one loadAverageMinute
fifteen loadAverageMinute
five loadAverageMinute
cached swap
free swap
in swap
out swap
total swap
maxFiles fileSys
usedFiles fileSys
usedFilePercent fileSys
usedPercent fileSys
used fileSys
total fileSys
rx network
tx network
numVCPUs general

Performance Insights Counters for Amazon RDS for MariaDB and MySQL

The following database counters are available with Performance Insights for Amazon RDS for MariaDB and MySQL.

Native Counters for RDS MariaDB and RDS MySQL

You can find definitions for these native metrics in Server Status Variables in the MySQL documentation.

Counter Type Unit
Com_analyze SQL Queries per second
Com_optimize SQL Queries per second
Com_select SQL Queries per second
Innodb_rows_deleted SQL Rows per second
Innodb_rows_inserted SQL Rows per second
Innodb_rows_read SQL Rows per second
Innodb_rows_updated SQL Rows per second
Select_full_join SQL Queries per second
Select_full_range_join SQL Queries per second
Select_range SQL Queries per second
Select_range_check SQL Queries per second
Select_scan SQL Queries per second
Slow_queries SQL Queries per second
Sort_merge_passes SQL Queries per second
Sort_range SQL Queries per second
Sort_rows SQL Queries per second
Sort_scan SQL Queries per second
Questions SQL Queries per second
Table_locks_immediate Locks Requests per second
Table_locks_waited Locks Requests per second
Innodb_row_lock_time Locks Milliseconds (average)
Aborted_clients Users Connections
Aborted_connects Users Connections
Threads_created Users Connections
Threads_running Users Connections
Innodb_data_writes IO Operations per second
Innodb_dblwr_writes IO Operations per second
Innodb_log_write_requests IO Operations per second
Innodb_log_writes IO Operations per second
Innodb_pages_written IO Pages per second
Created_tmp_disk_tables Temp Tables per second
Created_tmp_tables Temp Tables per second
Innodb_buffer_pool_pages_data Cache Pages
Innodb_buffer_pool_pages_total Cache Pages
Innodb_buffer_pool_read_requests Cache Pages per second
Innodb_buffer_pool_reads Cache Pages per second
Opened_tables Cache Tables
Opened_table_definitions Cache Tables
Qcache_hits Cache Queries

Non-Native Counters for Amazon RDS for MariaDB and MySQL

Non-native counter metrics are counters defined by Amazon RDS. A non-native metric can be a metric that you get with a specific query. A non-native metric also can be a derived metric, where two or more native counters are used in calculations for ratios, hit rates, or latencies.

Counter Type Description Definition
innodb_buffer_pool_hits Cache The number of reads that InnoDB could satisfy from the buffer pool. innodb_buffer_pool_read_requests - innodb_buffer_pool_reads
innodb_buffer_pool_hit_rate Cache The percentage of reads that InnoDB could satisfy from the buffer pool. 100 * innodb_buffer_pool_read_requests / (innodb_buffer_pool_read_requests + innodb_buffer_pool_reads)
innodb_buffer_pool_usage Cache

The percentage of the InnoDB buffer pool that contains data (pages).

Note

When using compressed tables, this value can vary. For more information, see the information about Innodb_buffer_pool_pages_data and Innodb_buffer_pool_pages_total in Server Status Variables in the MySQL documentation.

Innodb_buffer_pool_pages_data / Innodb_buffer_pool_pages_total * 100.0
query_cache_hit_rate Cache MySQL result set cache (query cache) hit ratio. Qcache_hits / (QCache_hits + Com_select) * 100
innodb_datafile_writes_to_disk IO The number of InnoDB data file writes to disk, excluding double write and redo logging write operations. Innodb_data_writes - Innodb_log_writes - Innodb_dblwr_writes
innodb_rows_changed SQL The total InnoDB row operations. db.SQL.Innodb_rows_inserted + db.SQL.Innodb_rows_deleted + db.SQL.Innodb_rows_updated
active_transactions Transactions The total active transactions. SELECT COUNT(1) AS active_transactions FROM INFORMATION_SCHEMA.INNODB_TRX
innodb_deadlocks Locks The total number of deadlocks. SELECT COUNT AS innodb_deadlocks FROM INFORMATION_SCHEMA.INNODB_METRICS WHERE NAME='lock_deadlocks'
innodb_lock_timeouts Locks The total number of deadlocks that timed out. SELECT COUNT AS innodb_lock_timeouts FROM INFORMATION_SCHEMA.INNODB_METRICS WHERE NAME='lock_timeouts'
innodb_row_lock_waits Locks The total number of row locks that resulted in a wait. SELECT COUNT AS innodb_row_lock_waits FROM INFORMATION_SCHEMA.INNODB_METRICS WHERE NAME='lock_row_lock_waits'

Performance Insights Counters for Amazon RDS for Oracle

The following database counters are available with Performance Insights for RDS Oracle.

Native Counters for RDS Oracle

You can find definitions for these native metrics in Statistics Descriptions in the Oracle documentation.

Note

For the CPU used by this session counter metric, the unit has been transformed from the native centiseconds to active sessions to make the value easier to use. For example, CPU send in the DB Load chart represents the demand for CPU. The counter metric CPU used by this session represents the amount of CPU used by Oracle sessions. You can compare CPU send to the CPU used by this session counter metric. When demand for CPU is higher than CPU used, sessions are waiting for CPU time.

Counter Type Unit
CPU used by this session User Active sessions. In this case, the unit has been transformed from the native centiseconds to active sessions to make the value easier to use.
SQL*Net roundtrips to/from client User Roundtrips per second
Bytes received via SQL*Net from client User Bytes per second
User commits User Commits per second
Logons cumulative User Logons per second
User calls User Calls per second
Bytes sent via SQL*Net to client User Bytes per second
User rollbacks User Rollbacks per second
Redo size Redo Bytes per second
Parse count (total) SQL Parses per second
Parse count (hard) SQL Parses per second
Table scan rows gotten SQL Rows per second
Sorts (memory) SQL Sorts per second
Sorts (disk) SQL Sorts per second
Sorts (rows) SQL Sorts per second
Physical read bytes Cache Bytes per second
DB block gets Cache Blocks per second
DBWR checkpoints Cache Checkpoints per minute
Physical reads Cache Reads per second
Consistent gets from cache Cache Gets per second
DB block gets from cache Cache Gets per second
Consistent gets Cache Gets per second

Performance Insights Counters for Amazon RDS PostgreSQL

The following database counters are available with Performance Insights for Amazon RDS PostgreSQL.

Native Counters for Amazon RDS PostgreSQL

You can find definitions for these native metrics in Viewing Statistics in the PostgreSQL documentation.

Counter Type Unit
blks_hit Cache Blocks per second
buffers_alloc Cache Blocks per second
buffers_checkpoint Checkpoint Blocks per second
checkpoint_sync_time Checkpoint Milliseconds per checkpoint
checkpoint_write_time Checkpoint Milliseconds per checkpoint
checkpoints_req Checkpoint Checkpoints per minute
checkpoints_timed Checkpoint Checkpoints per minute
maxwritten_clean Checkpoint Bgwriter clean stops per minute
deadlocks Concurrency Deadlocks per minute
blk_read_time IO Milliseconds
blks_read IO Blocks per second
buffers_backend IO Blocks per second
buffers_backend_fsync IO Blocks per second
buffers_clean IO Blocks per second
tup_deleted SQL Tuples per second
tup_fetched SQL Tuples per second
tup_inserted SQL Tuples per second
tup_returned SQL Tuples per second
tup_updated SQL Tuples per second
temp_bytes Temp Bytes per second
temp_files Temp Files per minute
active_transactions Transactions Transactions
blocked_transactions Transactions Transactions
max_used_xact_ids Transactions Transactions
xact_commit Transactions Commits per second
xact_rollback Transactions Rollbacks per second
numbackends User Connections
archived_count WAL Files per minute
archive_failed_count WAL Files per minute

Non-Native Counters for Amazon RDS PostgreSQL

Non-native counter metrics are counters defined by Amazon RDS. A non-native metric can be a metric that you get with a specific query. A non-native metric also can be a derived metric, where two or more native counters are used in calculations for ratios, hit rates, or latencies.

Counter Type Description Definition
checkpoint_sync_latency Checkpoint The total amount of time that has been spent in the portion of checkpoint processing where files are synchronized to disk. checkpoint_sync_time / (checkpoints_timed + checkpoints_req)
checkpoint_write_latency Checkpoint The total amount of time that has been spent in the portion of checkpoint processing where files are written to disk. checkpoint_write_time / (checkpoints_timed + checkpoints_req)
read_latency IO The time spent reading data file blocks by backends in this instance. blk_read_time / blks_read