[
https://issues.apache.org/jira/browse/HADOOP-18013?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17446597#comment-17446597
]
Steve Loughran commented on HADOOP-18013:
-----------------------------------------
We could add an abfs trash policy which skips the initial probe for the sauce
existing and swallows FNFEs.
Oh and any other exception to be passed up: don't wrap it with something that
loses the inner message.
This would seem a good time to add and the ability for a store to provide it's
own trash policy.
Today the trash policly factory looks at fs.trash.policy
but we could have it first look for fs.SCHEMA.trash.policy
* abfs would have the new one
* s3a could have one which just deletes rather than re-names. and it will work
on a per-bucket basis
This approach will let or provide a different policy for HDFS alongside object
stores
> ABFS: add resilient trash policy
> --------------------------------
>
> Key: HADOOP-18013
> URL: https://issues.apache.org/jira/browse/HADOOP-18013
> Project: Hadoop Common
> Issue Type: Improvement
> Components: fs/azure
> Affects Versions: 3.3.2
> Reporter: Steve Loughran
> Priority: Major
>
> Add a custom TrashPolicy for azure which
> * is patched in to the config automatically if "fs.trash.classname" is unset
> * checks the source for existing at all before trying anything
> * considers itself enabled even if the trash interval is 0
> * if interval is 0, just does a delete (issue: what about 90s timeouts)
> * downgrades exceptions/failures during rename. If anything is raised and the
> source dir isn't there, all is good.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
