Performance tuning for Ozone
Learn how to use the Ozone configuration properties to tune Ozone to work optimally on your cluster.
For performance optimization and to overcome known issues, Cloudera recommends the following
additional configurations:
Configuration Property | Value | Remarks |
---|---|---|
Maximum Process File Descriptors | 100,000 | |
Java Heap Size of Ozone Manager | 31GB | |
Java Heap Size of Storage Container Manager | 31GB | |
Java Heap Size of Recon | 31GB | Can be increased to 64GB for higher load. |
Java Heap Size of S3 Gateway | 31GB | |
Java Heap Size of DataNode | 31GB | |
Ozone Service Advanced Configuration Snippet (Safety Valve) for ozone-conf/ozone-site.xml |
|
Prometheus is an optional role and this configuration can be ignored if Prometheus is not being used. Disables Prometheus SPNEGO and uses Token Based Authentication. |
ozone.container.cache.size | 8192 | |
ozone.container.cache.lock.stripes | 8192 | |
hdds.datanode.du.factory.classname | org.apache.hadoop.hdds.fs.DedicatedDiskSpaceUsageFactory | |
hdds.container.report.interval | 10m | |
dfs.container.ratis.leader.pending.bytes.limit | 2GB | |
ozone.scm.datanode.pipeline.limit | 10 | If DataNodes have different number of disks, pick the highest number. |
ozone.scm.pipeline.owner.container.count | 10 | |
ozone.scm.pipeline.creation.auto.factor.one | FALSE | |
ozone.scm.container.placement.impl | org.apache.hadoop.hdds.scm.container.placement.algorithms.SCMContainerPlacementCapacity | To enable SCM to replicate containers based on capacity available instead of random allocation. |
ozone.recon.task.pipelinesync.interval | 120s | |
ozone.recon.task.missingcontainer.interval | 3600s | |
hdds.datanode.replication.work.dir | /tmp (is the default value) | The /tmp directory is used for staging transient files as part of data replication. Cloudera recommends a minimum of 50 GB size for the Datanode Replication Working Directory. Cloudera recommends you to use SSD for better performance during data replication. It can use the same SSD as datanode metadata. |
hdds.datanode.replication.work.dir | </path/to/custom/replication/work/dir> | To use a non-default directory as Datanode Replication Working Directory |
ozone.om.enable.filesystem.paths | true | |
scm.container.client.max.size | <number of active pipelines in the cluster> | |
scm.container.client.idle.threshold | 120s |