[
https://issues.apache.org/jira/browse/HBASE-12071?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14266573#comment-14266573
]
Nick Dimiduk commented on HBASE-12071:
--------------------------------------
What's the thinking on increasing the number of queues to 2?
{noformat}
- this.priorityExecutor =
- priorityHandlerCount > 0 ? new BalancedQueueRpcExecutor("Priority",
priorityHandlerCount,
- 1, maxQueueLength, conf, abortable) : null;
+ // Create 2 queues to help priorityExecutor be more scalable.
+ this.priorityExecutor = priorityHandlerCount > 0 ?
+ new BalancedQueueRpcExecutor("Priority", priorityHandlerCount, 2,
maxQueueLength) : null;
{noformat}
It's probably worth while turning priority into a protobuf enum. I don't know
if we can do that, given that RequestHeader.priority is already defined as a
uint32. The docs say valid enum value range is 32-bit signed int, so maybe we'd
have to deprecate the existing field? Someone who knows more about protobuf,
any thoughts here?
Other question: can we move the QosPriority annotations over to the
BlockingInterfaces? Maybe not since they're generated protobuf...
Otherwise, +1. You test all this out [~syuanjiang]? Working as expected?
> Separate out thread pool for Master <-> RegionServer communication
> ------------------------------------------------------------------
>
> Key: HBASE-12071
> URL: https://issues.apache.org/jira/browse/HBASE-12071
> Project: HBase
> Issue Type: Improvement
> Affects Versions: 2.0.0, 1.1.0
> Reporter: Sudarshan Kadambi
> Assignee: Stephen Yuan Jiang
> Labels: reliablity
> Fix For: 2.0.0, 1.1.0
>
> Attachments: HBASE-12071.v1-branch-1.patch,
> HBASE-12071.v2-master.patch, HBASE-12071.v2-master.patch,
> HBASE-12071.v3-master.patch, HBASE-12071.v4-master.patch,
> HBASE-12071.v5-master.patch
>
>
> Over in HBASE-12028, there is a discussion about the case of a RegionServer
> still being alive despite all its handler threads being dead. One outcome of
> this is that the Master is left hanging on the RS for completion of various
> operations - such as region un-assignment when a table is disabled. Does it
> make sense to create a separate thread pool for communication between the
> Master and the RS? This addresses not just the case of the RPC handler
> threads terminating but also long-running queries or co-processor executions
> holding up master operations.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
