[
https://issues.apache.org/jira/browse/HIVE-1818?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sushanth Sowmyan updated HIVE-1818:
-----------------------------------
Attachment: HIVE-1818-vs-1054860.patch
+Updated patch to work against current trunk.
+Also, made scopes ThreadLocal inside Metrics, which was necessary
+Renamed/removed some of the logStart* functions in HiveMetaStore so as to make
sure new contributions use the start* functions instead which also do the
required logging in addition to jmx tracking.
> Call frequency and duration metrics for HiveMetaStore via jmx
> -------------------------------------------------------------
>
> Key: HIVE-1818
> URL: https://issues.apache.org/jira/browse/HIVE-1818
> Project: Hive
> Issue Type: New Feature
> Components: Metastore
> Reporter: Sushanth Sowmyan
> Assignee: Sushanth Sowmyan
> Priority: Minor
> Attachments: HIVE-1818-vs-1054860.patch, HIVE-1818.patch
>
>
> As recently brought up in the hive-dev mailing list, it'd be useful if the
> HiveMetaStore had some sort of instrumentation capability so as to measure
> frequency of calls to various calls on the HiveMetaStore and the duration of
> time spent in these calls.
> There are already incrementCounter() and logStartFunction() /
> logStartTableFunction() ,etc calls in HiveMetaStore, and they could be
> refactored/repurposed to make calls that expose JMX MBeans as well. Or, a
> Metrics subsystem could be introduced which made calls to
> incrementCounter()/etc as a refactor.
> It might also be possible to specify a -D parameter that the Metrics
> subsystem could use to determine whether or not to be enabled, and if so, on
> to what port. And once we have the capability to instrument and expose
> MBeans, it might also be possible for other subsystems to also adopt and use
> this system.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
