-
Notifications
You must be signed in to change notification settings - Fork 5.5k
Monitoring
David Phillips edited this page Oct 5, 2016
·
1 revision
Presto exposes a huge number of metrics via JMX. This list provides a starting point for things you might want to monitor.
- Heap size:
java.lang:type=Memory:HeapMemoryUsage.used - Thread count:
java.lang:type=Threading:ThreadCount
- Active nodes:
com.facebook.presto.failureDetector:name=HeartbeatFailureDetector:ActiveCount - Active queries:
com.facebook.presto.execution:name=QueryManager:RunningQueries - Failed queries (all):
com.facebook.presto.execution:name=QueryManager:FailedQueries.FiveMinute.Count - Failed queries (internal):
com.facebook.presto.execution:name=QueryManager:InternalFailures.FiveMinute.Count - Failed queries (external):
com.facebook.presto.execution:name=QueryManager:ExternalFailures.FiveMinute.Count - Failed queries (user):
com.facebook.presto.execution:name=QueryManager:UserErrorFailures.FiveMinute.Count - Input data bytes:
com.facebook.presto.execution:name=TaskManager:InputDataSize.FiveMinute.Count - Input rows:
com.facebook.presto.execution:name=TaskManager:InputPositions.FiveMinute.Count - Queries started:
com.facebook.presto.execution:name=QueryManager:StartedQueries.FiveMinute.Count - Execution latency (P50):
com.facebook.presto.execution:name=querymanager:ExecutionTime.FiveMinutes.P50 - Free memory (general pool):
com.facebook.presto.memory:type=ClusterMemoryPool:name=general:FreeDistributedBytes - Input data rate (P90):
com.facebook.presto.execution:name=QueryManager:WallInputBytesRate.FiveMinutes.P90
- HDFS NameNode latency (P99):
com.facebook.presto.hive:name=NamenodeStats:ListLocatedStatus.Time.OneMinute.P99
Many of these metrics are actually a complex metric object such as a CounterStat that has a collection of related metrics. For example, InputPositions has InputPositions.TotalCount, InputPositions.OneMinute.Count, etc.