[
https://issues.apache.org/jira/browse/CASSANDRA-1355?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Folke Behrens updated CASSANDRA-1355:
-------------------------------------
Attachment: ftrace.000.traces.bz2
Very nice reads. You really talked me into writing my own tracing tool with my
very first ClassFileTransformer. Attached textfile shows one of Cassandra's
tests. The numbers on the left is the time in microseconds. 10 mio. means that
the thread never stopped or the method was somehow never left. The traces are
still raw but I already have a few optimization ideas, like time accumulation,
filtering out fast methods and estimating the time where it couldn't be
calculated.
There are probably tons of profilers with selective tracing out there but this
was really fun. Nevertheless, my original point still stands, methods should be
logged before they leave.
> Log RPC exceptions and timing data
> ----------------------------------
>
> Key: CASSANDRA-1355
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1355
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Folke Behrens
> Priority: Minor
> Fix For: 0.7.1
>
> Attachments: 0001-RPCLogger.patch,
> 0002-conditional-wrapping.patch.txt, ftrace.000.traces.bz2
>
>
> Attached patch introduces a logger that sits between Cassandra.Processor and
> CassandraServer. It logs exceptions before they get send to the client and it
> can output timing and tracing information. Pretty useful.
> This extra layer would actually be a good place for authentication and
> authorization as well.
> Thoughts?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
