[jira] [Updated] (OAK-2501) add more trace-logging to mongoMk to debug performance issues

["Stefan Egli (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/OAK-2501?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Stefan Egli updated OAK-2501:
-----------------------------
    Attachment: oak.2501.patch

Attaching oak.2501.patch applicable to current 1.0 branch, containing PerfLogger which is used for a couple of key mongoMk operations and should help trace performance issues. Two different modes are available: when log level is set to DEBUG then only operations taking longer than 1ms are logged, only at the 'end' of the operation. When log level is set to TRACE, then 'start' is also logged (if a log message is provided at the particular location, not at all of them) and 'end' is always logged. [~mreutegg], [~chetanm], can you pls review & apply? Thx!

> add more trace-logging to mongoMk to debug performance issues
> -------------------------------------------------------------
>
>                 Key: OAK-2501
>                 URL: https://issues.apache.org/jira/browse/OAK-2501
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>          Components: core
>    Affects Versions: 1.0.11
>            Reporter: Stefan Egli
>             Fix For: 1.0.12
>
>         Attachments: oak.2501.patch
>
>
> Debugging performance issues within mongoMk is currently somewhat more challenging that it would have to - IMO because there is no easy logger that can be turned on to get some performance numbers. Ie currently for figuring out what's going on in mongoMk you can set the log level to TRACE for the following loggers:
> {code}
> org.apache.jackrabbit.oak.plugins.observation.EventGenerator
> org.apache.jackrabbit.oak.plugins.document.mongo.MongoDocumentStore
> org.apache.jackrabbit.oak.plugins.document.NodeDocument
> org.apache.jackrabbit.oak.plugins.document.DocumentNodeStore 
> {code}
> while you might still not catch all interesting and crucial details.
> So the suggestion is to add a dedicated logger which is used at a couple of neuralgic locations (eg in observation event generation, cache handling, diffChildren, findUncached etc)



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)