Hi Biao,
Thanks for getting back to me.
Here's the command I used:
/usr/bin/flink run-application -t yarn-application \
-Dtaskmanager.numberOfTaskSlots=1 \
-Djobmanager.memory.process.size=4096m \
-Dtaskmanager.memory.process.size=4096m \
-Dyarn.application.name=backup-enriched-events-r0 \
-c com.xyz.source.SourceDataStream \
/tmp/lib-event-platform-core-7c7cc8ca9fc.jar \
--datastreamName=backup-enriched-events-r0 --environment=dev
And with the following configuration in flink config:
yarn.per-job-cluster.include-user-jar: FIRST
classloader.resolve-order: parent-first
Thanks
Leon
On Wed, Nov 30, 2022 at 10:14 PM Biao Geng <[email protected]> wrote:
> Hi Leon,
>
> Can you share your full command for submission?
>
>
> Best,
> Biao Geng
>
> Leon Xu <[email protected]> 于2022年12月1日周四 06:27写道：
>
>> Hi Flink Users,
>>
>> We ran into java.lang.ClassCastException after moving the flink job from
>> session mode to application mode.
>>
>>
>> *java.lang.ClassCastException: class [B cannot be cast to class
>> java.lang.String ([B and java.lang.String are in module java.base of loader
>> 'bootstrap') at
>> com.xyz.common.io.sink.HivePartitionedBucketAssigner.getBucketId(HivePartitionedBucketAssigner.java:35)*
>>
>> Here *HivePartitionedBucketAssigner* is a template class:
>> class HivePartitionedBucketAssigner<T> implements BucketAssigner<T,
>> String>
>>
>>
>> We tried disabling the invert class loading based on this doc
>> <https://nightlies.apache.org/flink/flink-docs-master/docs/ops/debugging/debugging_classloading/#x-cannot-be-cast-to-x-exceptions>but
>> it didn't help. In session-mode it works fine. So I wonder what's different
>> between session-mode and application mode in terms of class loading? And
>> what would be a solution for this situation?
>> Our setup:
>> 1. Flink version: 1.12.7
>> 2. Java: JDK 11
>>
>>
>> Thanks
>> Leon
>>
>
