hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Riley Patterson (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-4117) Slow Query Log
Date Tue, 16 Aug 2011 23:45:27 GMT

    [ https://issues.apache.org/jira/browse/HBASE-4117?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13086045#comment-13086045
] 

Riley Patterson commented on HBASE-4117:
----------------------------------------

This isn't really meant to enable gathering of status information. It is a bunch of historical
information which is highly related to other information in the main log. Without the context
of log events regarding compactions, region reassignments, etc. that we gain by placing it
in the main log, we lose most of the utility. It can also provide context to other log events
that may have been caused by RPC handlers getting locked up on slow queries.

We could additionally log this information somewhere in a monitor in memory or in its own
separate log and open this up to RPC access, but it would duplicate a lot of historical information
which would likely see infrequent use. I don't think there's any more reason to do this for
this particular feature than for most other events logged in our main log file.

Note that I'm also finishing up work on an initial implementation of HBASE-4057, which deals
directly with status information for the calls currently being serviced by RPC handlers and
enables exposure of the contents of getFingerprint and getDetails in a more general way. This
information resides in the TaskMonitor singleton, which can then be exposed any way you like.
I think this design makes a lot of sense; historical information goes in the main log, while
current status information can be queried from a clean interface and exposed any way you like.

> Slow Query Log
> --------------
>
>                 Key: HBASE-4117
>                 URL: https://issues.apache.org/jira/browse/HBASE-4117
>             Project: HBase
>          Issue Type: New Feature
>          Components: ipc
>            Reporter: Riley Patterson
>            Assignee: Riley Patterson
>            Priority: Minor
>              Labels: client, ipc
>         Attachments: HBASE-4117.patch
>
>
> Produce log messages for slow queries. The RPC server will decide what is slow based
on a configurable "warn response time" parameter. Queries designated as slow will then output
a "response too slow" message followed by a fingerprint of the query, and a summary limited
in size by another configurable parameter (to limit log spamming).

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message