[
https://issues.apache.org/jira/browse/HBASE-13054?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14328139#comment-14328139
]
Elliott Clark commented on HBASE-13054:
---------------------------------------
{code}Trace.startSpan("MemStoreScanner").close();{code}
This should be an annotation.
{code}traceScope.getSpan().addTimelineAnnotation("Waiting for row lock");{code}
No need to add that annotation since that's the purpose of the span.
{code}if (traceScope != null)
traceScope.getSpan().addTimelineAnnotation("Acquired row lock");{code}
I'd rather just annotate the failure case since that will be the more un-usual
case.
> Provide more tracing information for locking/latching events.
> -------------------------------------------------------------
>
> Key: HBASE-13054
> URL: https://issues.apache.org/jira/browse/HBASE-13054
> Project: HBase
> Issue Type: Improvement
> Reporter: Rajeshbabu Chintaguntla
> Assignee: Rajeshbabu Chintaguntla
> Fix For: 2.0.0, 1.0.1, 1.1.0, 0.98.11
>
> Attachments: HBASE-13054.patch, HBASE-13054_v2.patch
>
>
> Currently not much tracing information available for locking and latching
> events like row level locking during do mini batch mutations, region level
> locking during flush, close and so on. It will be better to add the trace
> information for such events so that it will be useful for finding time spent
> on locking and waiting time on locks while analyzing performance issues in
> queries using trace information.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
