[ 
https://issues.apache.org/jira/browse/IMPALA-8674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17045992#comment-17045992
 ] 

Tim Armstrong commented on IMPALA-8674:
---------------------------------------

[~rizaon] thanks for picking this up, it is really confusing so would be good 
to fix. I took a look and it looks to me that we essentially need an additional 
bit of information about whether Coordinator::FilterState::Disable() was called 
because the filter was complete and sent out, or whether it wasn't sent out. It 
probably makes sense to look at each call to Disable() and decide how that 
should be represented in the filter table. 

I think your proposed solution probably does all that, but just describing my 
thought process.

> REMOTE runtime filters are always shown as disabled in the final filter 
> routing table
> -------------------------------------------------------------------------------------
>
>                 Key: IMPALA-8674
>                 URL: https://issues.apache.org/jira/browse/IMPALA-8674
>             Project: IMPALA
>          Issue Type: Bug
>          Components: Backend
>    Affects Versions: Impala 3.2.0
>            Reporter: Tim Armstrong
>            Assignee: Riza Suminto
>            Priority: Major
>              Labels: observability, ramp-up
>
> {noformat}
> Final filter table:
> ID  Src. Node  Tgt. Node(s)     Target type  Partition filter  Pending 
> (Expected)  First arrived  Completed   Enabled
> ----------------------------------------------------------------------------------------------------------------------
>  1          6          0, 1  REMOTE, REMOTE      false, false              0 
> (93)       50s381ms   53s566ms     false
>  0          6             0          REMOTE              true              0 
> (93)       50s177ms   52s336ms     false
> {noformat}
> The issue is that the filter is disabled after the remote filter is sent out 
> https://github.com/apache/impala/blob/95a1da2/be/src/runtime/coordinator.cc#L987.
>  The code seems to really conflate releasing of resources with filters that 
> were logically disabled.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to