[
https://issues.apache.org/jira/browse/HBASE-18662?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16160056#comment-16160056
]
Chia-Ping Tsai commented on HBASE-18662:
----------------------------------------
It is hard to choose a appropriate value for
{{hbase.client.max.perserver.tasks}}, because that is a tradeoff between the
number of concurrency threads and server loading. If we have to pick up a
default value for {{hbase.client.max.perserver.tasks}}, the 2 will be more
suitable than 5. The {{hbase.client.max.perserver.tasks}} works only when
someone call AsyncProcess#submit with
AsyncProcessTask.SubmittedRows.AT_LEAST_ONE. After HBASE-18500, only
BufferedMutatorImpl, which is not thread-safe, applies the function. That is to
say, the 5 means that we may invoke 5 threads to serve 1 user thread when user
try to submit data via BMI. Personally, that is too radical to be a *default
value*.
> The default values for many configuration items in the code are not
> consistent with hbase-default.xml
> -----------------------------------------------------------------------------------------------------
>
> Key: HBASE-18662
> URL: https://issues.apache.org/jira/browse/HBASE-18662
> Project: HBase
> Issue Type: Improvement
> Reporter: Yun Zhao
> Assignee: Yun Zhao
> Priority: Minor
> Fix For: 2.0.0-alpha-3
>
> Attachments: HBASE-18662.master.002.patch
>
>
> Such as
> {code}
> hbase.ipc.server.callqueue.handler.factor
> hbase.regionserver.logroll.errors.tolerated
> hbase.regionserver.region.split.policy
> zookeeper.session.timeout
> hbase.client.retries.number
> hbase.client.max.perserver.tasks
> hbase.client.keyvalue.maxsize
> hbase.normalizer.period
> hbase.hstore.blockingStoreFiles
> hbase.snapshot.restore.take.failsafe.snapshot
> hbase.lease.recovery.dfs.timeout
> hbase.rest.filter.classes
> hbase.http.max.threads
> {code}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
