[
https://issues.apache.org/jira/browse/DRILL-8448?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17750360#comment-17750360
]
James Turton edited comment on DRILL-8448 at 8/2/23 2:26 PM:
-------------------------------------------------------------
Thanks for reporting this. There a few Drill Jira issues that talk about
queries getting stuck in CANCELLATION_REQUESTED, e.g. DRILL-5500. I believe
that any failure that leaves one or more fragments waiting for a lock (monitor)
that never gets released can result in a query stuck in this state. Such
failures can happen as a result of a failed HBase pstore access but also in
other ways as can be seen in the other Jira issues.
[~volodymyr] [~cgivre] one general way to fix this might be to set a query
cancellation timeout after which any fragment threads that are still running
(or blocked) get forcibly interrupted - has anything like this ever been
discussed?
was (Author: dzamo):
Thanks for reporting this. There a few Drill Jira issues that talk about
queries getting stuck in CANCELLATION_REQUESTED, e.g. DRILL-5500. I believe
that any failure that leaves one or more fragments waiting for a lock (monitor)
that never gets released can result in a query stuck in this state. Such
failures can happen as a result of a failed HBase pstore access but also in
other ways as can be seen in the other Jira issues.
[~volodymyr] [~cgivre] one way to fix this might be to set a query cancellation
timeout after which any fragment threads that are still running (or blocked)
get forcibly interrupted - has anything like this ever been discussed?
> Query cancellation not possible with hbase persistent store
> -----------------------------------------------------------
>
> Key: DRILL-8448
> URL: https://issues.apache.org/jira/browse/DRILL-8448
> Project: Apache Drill
> Issue Type: Bug
> Affects Versions: 1.20.3
> Reporter: Christian Pfarr
> Priority: Minor
> Attachments: 1b3dcd7e-60b4-069f-0c53-c1ee168b2d9c.json,
> 1b3dcd7e-60b4-069f-0c53-c1ee168b2d9c.log,
> 1b419c0e-0018-3775-794b-77513f64df56.json,
> 1b419c0e-0018-3775-794b-77513f64df56.log, image-2023-08-02-08-47-40-413.png
>
>
> When we cancel a running query it is hanging in the state
> CANCELLATION_REQUESTED and stays in the list of running queries via webui.
> !image-2023-08-02-08-47-40-413.png!
> In the logs we see that these queries has failed because of an error with our
> hbase persistent store. Maybe its just a bug in the UI?
> I´ve attached the profiles and the logs for this cancellation requests.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
