[Feature] Improve the tryCommitOnce behavior in FileStoreCommitImpl

[xiangyuf]

Search before asking

• I searched in the issues and found nothing similar.

Motivation

When using dedicated compactions in production, we've found the write only job and compact job will failover every 2 or 3 days even if the remote filesystem support atomic rename operation.

The main cause is the FileAlreadyExistsException:

Checking with recent Hadoop API Rename implementation, we found the rename api will return FileAlreadyExistsException for rename api instead of false by default.
https://github.com/apache/hadoop/blob/branch-3.3.6/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSDirRenameOp.java

IMHO, this can be improved by catch certain exceptions in tryCommitOnce and return false to upper caller.

Solution

No response

Anything else?

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!

[xiangyuf]

Hi @JingsongLi, WDYT about this?

[tsreaper]

Could you explain what kinds of exceptions you would like to catch, and where you would like to catch them?