Taraka Rama Rao Lethavadla created HIVE-29283:
-------------------------------------------------
Summary: Display reason/txn id that is causing ready for cleaning
state of a compaction
Key: HIVE-29283
URL: https://issues.apache.org/jira/browse/HIVE-29283
Project: Hive
Issue Type: Improvement
Components: Metastore, Standalone Metastore
Reporter: Taraka Rama Rao Lethavadla
*Problem Statement:*
We see compaction requests are showing 'ready for cleaning' state. And
sometimes, it will not finish and stuck there. There could be multiple reasons
like a {*}long running transaction{*}, Obsolete directories were already
cleaned, Cleaner stopped the attempt to clean for this specific compaction
request etc.
So the specific reason about why a compaction is in ready for cleaning can only
be known after looking at the logs which is very difficult and would take a lot
of time to get to a conclusion on the RCA
Another case is where Initiator not able to initiate a compaction request for
some reason like file not found etc.. We can capture this one as well and
display
*Proposal:*
Display the reason for 'ready for cleaning/failed to initiate' state of a
compaction request.
If it is due to long running txn, we should display that txn id, so that if
that txn happens to be stale/invalid/unwanted etc, then that can be aborted so
that compaction can continue
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
