[
https://issues.apache.org/jira/browse/HBASE-15931?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yu Li updated HBASE-15931:
--------------------------
Attachment: HBASE-15931.patch
Here's the initial patch.
Introduce two thresholds for the detailed logging:
{{hbase.client.threshold.log.details}} default to 10 which means to start
logging regionservers processing the in-progress tasks when 10 or less task
left, and hard-coded {{THRESHOLD_TO_LOG_REGION_DETAILS}} which means to start
logging regions holding requests when there're only 2 or less tasks left.
Also add back logging for table name so we could more easily locate the
regionserver processing the long-tail request.
> Add log for long-running tasks in AsyncProcess
> ----------------------------------------------
>
> Key: HBASE-15931
> URL: https://issues.apache.org/jira/browse/HBASE-15931
> Project: HBase
> Issue Type: Improvement
> Reporter: Yu Li
> Assignee: Yu Li
> Attachments: HBASE-15931.patch
>
>
> Currently if there's any long-tail tasks in a multi-action request like
> triggered by {{BufferedMutatorImpl#backgroundFlushCommits}}, we could see
> logging message like below
> {noformat}
> 2016-05-31 09:36:55,461 INFO [Thread-16]
> org.apache.hadoop.hbase.client.AsyncProcess: #28, waiting for some tasks to
> finish. Expected max=0, tasksInProgress=1
> {noformat}
> but there's no way to know detail of this long-tail, such as which
> table/region it's accessing. This JIRA aims at adding such log for easier
> debugging.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
