[
https://issues.apache.org/jira/browse/HBASE-20499?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Duo Zhang resolved HBASE-20499.
-------------------------------
Fix Version/s: 2.5.0
Resolution: Fixed
Cherry-picked to branch-2 and branch-2,5,
> Replication/Priority executors can use specific max queue length as default
> value instead of general maxQueueLength
> -------------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-20499
> URL: https://issues.apache.org/jira/browse/HBASE-20499
> Project: HBase
> Issue Type: Improvement
> Components: rpc
> Reporter: Nihal Jain
> Assignee: Nihal Jain
> Priority: Minor
> Fix For: 2.5.0, 3.0.0-alpha-1
>
> Attachments: 20499.master.002.patch, HBASE-20499.master.001.patch,
> HBASE-20499.master.002.patch
>
>
> {{In *SimpleRpcScheduler*'s constructor, instead of the using
> *maxQueueLength* as default value for
> *replicationExecutor*/*replicationExecutor*'s max queue length:}}
> {code:java}
> int maxQueueLength = conf.getInt(RpcScheduler.IPC_SERVER_MAX_CALLQUEUE_LENGTH,
> handlerCount * RpcServer.DEFAULT_MAX_CALLQUEUE_LENGTH_PER_HANDLER);
> int maxPriorityQueueLength =
> conf.getInt(RpcScheduler.IPC_SERVER_PRIORITY_MAX_CALLQUEUE_LENGTH,
> maxQueueLength);
> .
> .
> this.replicationExecutor = replicationHandlerCount > 0 ? new
> FastPathBalancedQueueRpcExecutor(
> "replication.FPBQ", replicationHandlerCount,
> RpcExecutor.CALL_QUEUE_TYPE_FIFO_CONF_VALUE,
> maxQueueLength, priority, conf, abortable) : null;{code}
> [See
> SimpleRpcScheduler|https://github.com/apache/hbase/blob/96ed407c691ac0686fb14cdcd8680d1849e24ae8/hbase-server/src/main/java/org/apache/hadoop/hbase/ipc/SimpleRpcScheduler.java#L97]
> We can do the following:
> {code:java}
> int maxQueueLength = conf.getInt(RpcScheduler.IPC_SERVER_MAX_CALLQUEUE_LENGTH,
> handlerCount * RpcServer.DEFAULT_MAX_CALLQUEUE_LENGTH_PER_HANDLER);
> int maxPriorityQueueLength =
> conf.getInt(RpcScheduler.IPC_SERVER_PRIORITY_MAX_CALLQUEUE_LENGTH,
> priorityHandlerCount * RpcServer.DEFAULT_MAX_CALLQUEUE_LENGTH_PER_HANDLER);
> .
> .
> int maxQueueLengthForReplication =
> conf.getInt(RpcScheduler.IPC_SERVER_MAX_CALLQUEUE_LENGTH,
> replicationHandlerCount * RpcServer.DEFAULT_MAX_CALLQUEUE_LENGTH_PER_HANDLER);
> .
> .
> this.replicationExecutor = replicationHandlerCount > 0 ? new
> FastPathBalancedQueueRpcExecutor(
> "replication.FPBQ", replicationHandlerCount,
> RpcExecutor.CALL_QUEUE_TYPE_FIFO_CONF_VALUE,
> maxQueueLengthForReplication , priority, conf, abortable) : null;
> {code}
>
> Also, we can make the maximum replication call queue length configurable,
> similar to general and priority call queue length.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
