cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Ellis (JIRA)" <j...@apache.org>
Subject [jira] [Resolved] (CASSANDRA-8380) Only record trace if query exceeds latency threshold.
Date Mon, 01 Dec 2014 07:04:12 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-8380?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Jonathan Ellis resolved CASSANDRA-8380.
---------------------------------------
    Resolution: Won't Fix

You'll need to post-process the traces.  Each node involved in a traced query inserts its
activity into the trace tables directly, so there's no way to tell it "never mind."

(We could delete it on the coordinator post facto but at that point we're not doing any better
than you can do client side anyway.)

> Only record trace if query exceeds latency threshold.
> -----------------------------------------------------
>
>                 Key: CASSANDRA-8380
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-8380
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Matt Stump
>            Priority: Critical
>
> Probabilistic trace isn't very useful because you're typically trying to only find the
badly performing queries which may only be .01% of requests. I would like to enable probabilistic
trace on a sample of queries, but then only record the trace if the request exceeded a time
threshold. This would allow us to more easily isolate performance problems.



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

Mime
View raw message