You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues-all@impala.apache.org by "Sahil Takiar (Jira)" <ji...@apache.org> on 2020/03/21 19:24:00 UTC

[jira] [Commented] (IMPALA-9233) Add impalad level metrics for query retries

    [ https://issues.apache.org/jira/browse/IMPALA-9233?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17064032#comment-17064032 ] 

Sahil Takiar commented on IMPALA-9233:
--------------------------------------

It would be good to have a metric that tracks the number of in-flight retried queries and fragments that are running. This can give users a sense a how much additional load on the system retries are adding.

We have similar Web UI metrics for in-flight queries.

> Add impalad level metrics for query retries
> -------------------------------------------
>
>                 Key: IMPALA-9233
>                 URL: https://issues.apache.org/jira/browse/IMPALA-9233
>             Project: IMPALA
>          Issue Type: Sub-task
>          Components: Backend
>            Reporter: Sahil Takiar
>            Priority: Major
>
> It would nice to have some impalad level metrics related to query retries. This would help answer questions like - how often are queries retried? how often are the retries actually successful? If queries are constantly being retried, then there is probably something wrong with the cluster.
> Some possible metrics to add:
>  * Query retry rate (the rate at which queries are retried)
>  ** This can be further divided by retry “type” - e.g. what caused the retry
>  ** Potential categories would be:
>  *** Queries retried due to failed RPCs
>  *** Queries retried due to faulty disks
>  *** Queries retried due to statestore detection of cluster membership changes
>  * A metric that measures how often query retries are actually successful (e.g. if a query is retried, does the retry succeed, or does it just fail again)
>  ** This can help users determine if query retries are actually helping, or just adding overhead (e.g. if retries always fail then something is probably wrong)



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscribe@impala.apache.org
For additional commands, e-mail: issues-all-help@impala.apache.org