[
https://issues.apache.org/jira/browse/PIG-5456?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Koji Noguchi updated PIG-5456:
------------------------------
Attachment: pig-5456-v02.patch
> log4j-1.2.17.jar was coming from stale zookeeper. Will create a new Jira to
> update the dependency.
Created PIG-5457
> As for how to skip reload4j
>
One option I considered was to move the reload4j to a different directory and
only pick it up for non-spark3 jobs. This may work if the way to start up pig
was only from bin/pig or unit/e2e tests. However, given we don't know if
users have such custom startup script(s), taking another approach. Leaving
the reload4j jar in the same location but explicitly skipping it from bin/pig
and build.xml(unit) tests. This way, only Pig-on-spark jobs are affected
leaving the rest untouched. (pig-5456-v02.patch)
> Upgrade Spark to 3.4.3
> ----------------------
>
> Key: PIG-5456
> URL: https://issues.apache.org/jira/browse/PIG-5456
> Project: Pig
> Issue Type: Improvement
> Components: spark
> Reporter: Koji Noguchi
> Assignee: Koji Noguchi
> Priority: Major
> Fix For: 0.19.0
>
> Attachments: pig-5456-v01.patch, pig-5456-v02.patch
>
>
> Major blocker for upgrading to Spark 3.4.3 was Spark started using log4j2.
> Simple upgrade failing a lot of tests with
> {noformat}
> java.lang.VerifyError: class org.apache.log4j.bridge.LogEventAdapter
> overrides final method getTimeStamp.()J {noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
