[
https://issues.apache.org/jira/browse/IMPALA-12022?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17927754#comment-17927754
]
Quanlong Huang commented on IMPALA-12022:
-----------------------------------------
I think we can resolve this now. IMPALA-12024 and IMPALA-12443 add catalog
timeline in the query profiles. Time waiting for metastoreDdlLock is shown in
the timeline, e.g.
{noformat}
Catalog Server Operation: 26s560ms
- Got metastoreDdlLock: 163.977us (163.977us)
- Got Metastore client: 166.339us (2.362us)
- Got current Metastore event id 8355270: 26s494ms (26s494ms)
- Created table in Metastore: 26s558ms (63.507ms)
- Fetched event batch from Metastore: 26s559ms (1.155ms)
- Created table in catalog cache: 26s560ms (1.164ms)
- DDL finished: 26s560ms (65.538us) {noformat}
> Keep metrics about wait times for metastoreDdlLock_
> ---------------------------------------------------
>
> Key: IMPALA-12022
> URL: https://issues.apache.org/jira/browse/IMPALA-12022
> Project: IMPALA
> Issue Type: Improvement
> Components: Catalog
> Affects Versions: Impala 4.3.0
> Reporter: Joe McDonnell
> Priority: Major
>
> CatalogOpExecutor's metastoreDdlLock_ is used to synchronize operations to
> create/drop objects in HMS with the local state of the catalog. If the HMS is
> slow for any reason, threads doing create/drop operations will end up
> blocking on metastoreDdlLock_. It would be useful to keep track of the time
> spent waiting on this lock, as a high value could indicate HMS slowness.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
