[jira] [Updated] (CASSANDRA-12626) LWT contention appears to be under reported

["Christopher Batey (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CASSANDRA-12626?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Christopher Batey updated CASSANDRA-12626:
------------------------------------------
    Description: 
I was surprised how little contention was being reported and looking at the code we miss out reporting contention when there is a timeout at the prepare phase.

If this {{StorageProxy.beginAndRepairPaxos}} throws then the contentions count isn't added to the histogram.

  was:
I was surprised how little contention was being reported and looking at the code we miss out reporting contention when there is a timeout at the prepare phase.

If this {{StorageProxy.beginAndRepairPaxos}} throws then the contentions variable isn't added to the metric.

The current way we do it means that we only update the counter once in the finally but given the expense of LWTs / the fact we also sleep after contention perhaps we should just increment it in the loop.


> LWT contention appears to be under reported
> -------------------------------------------
>
>                 Key: CASSANDRA-12626
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-12626
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Observability
>            Reporter: Christopher Batey
>            Assignee: Christopher Batey
>            Priority: Minor
>             Fix For: 3.x
>
>
> I was surprised how little contention was being reported and looking at the code we miss out reporting contention when there is a timeout at the prepare phase.
> If this {{StorageProxy.beginAndRepairPaxos}} throws then the contentions count isn't added to the histogram.



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