hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-2333) Automated anomoly report and anonymous usage statistics collection
Date Wed, 17 Mar 2010 03:29:27 GMT

    [ https://issues.apache.org/jira/browse/HBASE-2333?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12846279#action_12846279
] 

stack commented on HBASE-2333:
------------------------------

I like this idea.  Maybe we could have it  sent to a table in the mozilla cluster.

> Automated anomoly report and anonymous usage statistics collection
> ------------------------------------------------------------------
>
>                 Key: HBASE-2333
>                 URL: https://issues.apache.org/jira/browse/HBASE-2333
>             Project: Hadoop HBase
>          Issue Type: New Feature
>            Reporter: Andrew Purtell
>            Priority: Minor
>
> Collection of anonymous usage data from users willing to participate can help the project
in several ways:
> - Characterization of typical workloads
> - Long term trending of various performance metrics across releases
> This could be done by having the master collect information from the region servers and
itself over a 24 hour period then send a report to a configured URL, some web service up on
*.hbase.org. The information would be anonymized according to detail put up on the wiki. Each
master would identify itself via a GUID built in part from MAC address. For the above items,
only aggregated statistics are interesting, number of ops/hour/server, where ops are such
things as get, put, scanner.next, split, compact, etc. At the same time, sample HDFS metrics
and system metrics (cpu, ram, wio) over the same interval. 
> Later some more involved reporting activities can be considered:
> - Trigger based autonomous switch to DEBUG mode and log collection for some period --
like automated crash reports -- given failure or stress indications.
> For this type of activity, table names, keys, and server names would be replaced with
hashes of them, sufficient for event correlation for debugging but using cryptographically
strong one way functions to fully obfuscate details of the application. 

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message