[
https://issues.apache.org/jira/browse/CASSANDRA-10983?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15390095#comment-15390095
]
Sharvanath Pathak commented on CASSANDRA-10983:
-----------------------------------------------
[~jbellis] Our use case was to look at the system, and say what queries have
been the culprits. Tracing doesn't allow us to look at the history. I agree
that the metric doesn't look ideal for this, but it allows us to easily
visualize the queries that are bad and compare amongst them. Any other option
that addresses both use cases?
> Metrics for tracking offending queries
> --------------------------------------
>
> Key: CASSANDRA-10983
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10983
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Sharvanath Pathak
> Labels: github-import
> Fix For: 2.1.x
>
>
> I have seen big GC pauses leading to nodes being marked DOWN in our cluster.
> The most common issue is someone, would add a large range scan and it would
> be difficult to pin-point the specific query. I have added a mechanism to
> account the memory allocation for a specific query. In order to allow
> aggregates over a period I added a metric as well. Attached is the diff.
> I was wondering if something like this would be interesting for more general
> audience. There are some things which need to be fixed for proper release.
> For instance, Cleaning up existing metrics on server restart. However, just
> wanted to check before that if something like this would be useful for others.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
