[
https://issues.apache.org/jira/browse/HIVE-27097?focusedWorklogId=851816&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-851816
]
ASF GitHub Bot logged work on HIVE-27097:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 20/Mar/23 16:01
Start Date: 20/Mar/23 16:01
Worklog Time Spent: 10m
Work Description: wecharyu commented on code in PR #4076:
URL: https://github.com/apache/hive/pull/4076#discussion_r1142356918
##########
standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/RetryingHMSHandler.java:
##########
@@ -227,6 +237,23 @@ public Result invokeInternal(final Object proxy, final
Method method, final Obje
}
}
+ private boolean isRecoverableException(Throwable t) {
Review Comment:
The RPC flow in HMS looks like:
**RetryingMetaStoreClient <---> RetryingHMSHandler <---> DBMS**
Such recoverable exceptions always occur between `RetryingHMSHandler` and
`DBMS`, I think it's RetryingHMSHandler's duty to handle this retry strategy.
And this can also reduce the reconnection from metastore client to metastore
server.
Issue Time Tracking
-------------------
Worklog Id: (was: 851816)
Time Spent: 2h 50m (was: 2h 40m)
> Improve the retry strategy for Metastore client and server
> ----------------------------------------------------------
>
> Key: HIVE-27097
> URL: https://issues.apache.org/jira/browse/HIVE-27097
> Project: Hive
> Issue Type: Improvement
> Components: Hive
> Affects Versions: 4.0.0-alpha-2
> Reporter: Wechar
> Assignee: Wechar
> Priority: Major
> Labels: pull-request-available
> Time Spent: 2h 50m
> Remaining Estimate: 0h
>
> *Background*
> Hive provides *{{RetryingMetaStoreClient}}* and *{{RetryingHMSHandler}}* to
> do retry when thrift request failed:
> * RetryingMetaStoreClient will retry for *thrift related exception* and some
> *MetaException*
> * RetryingHMSHandler will retry for all {*}JDOException{*} or
> *NucleusException*.
> *Motivation*
> Current retry mechanism will lead to many unnecessary retries in both client
> and server. To simplify the process, we introduce following retry mechanism:
> * Client side only concerns the error of communication, i.e.,
> {*}TTransportException{*}.
> * Server side can skip some exceptions which always turn to fail even with
> retry, like {*}SQLIntegrityConstraintViolationException{*}.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
