[ 
https://issues.apache.org/jira/browse/OAK-6935?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16251180#comment-16251180
 ] 

Vikas Saurabh commented on OAK-6935:
------------------------------------

[~amjain], [~chetanm], while working on this and JCR-4213, it seemed that we 
might be ok to ignore (log at DEBUG) for any {{DataStoreException}}. The 
rationale being, that:
* this feature was always intended as best-effort-basis
* if there's a genuine error, even then it doesn't affect the functionality of 
this feature (other that, of course, unable to delete some blobs that it 
"wanted" to delete)
* current logging is just a WARN to the user - but, there is no way that the 
user/system can resume deletion of what has already been missed due to exception

So, I think we can probably close JCR-4213 as won't fix or later and fix this 
one with switching log.warn to log.debug for DataStoreExceptions. What do you 
guys think?

> Active deletion logs warn messages when it tries to delete blobs already 
> purged by DSGC
> ---------------------------------------------------------------------------------------
>
>                 Key: OAK-6935
>                 URL: https://issues.apache.org/jira/browse/OAK-6935
>             Project: Jackrabbit Oak
>          Issue Type: Bug
>          Components: lucene
>            Reporter: Vikas Saurabh
>            Assignee: Vikas Saurabh
>            Priority: Minor
>
> In cases when BlobGC deleted the blobs which active delete logic later came 
> around to purge., active deletion logs a warn message. An example timeline 
> would be:
> * t1 -> index blob gets deleted (active delete logic notes it down)
> * t2 -> compaction removes ref to this blob from node store
> * t3 (t1 + 24 hours) -> blob gc runs and collects the blob
> * t4 -> active gc tries to purge an already deleted blob
> (note, active delete runs between t1 and t3 would not delete the blobs as it 
> hasn't be 24 hours since blob deletion).
> While this would happen for a small set of blobs (that are unfortunate to get 
> compacted and cleaned up by DSGC before active deletion comes around) - but, 
> the warn log is annoying. It would be useful to detect this case and not log 
> warn on failure to delete for these.
> /cc [~amjain]



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to