[
https://issues.apache.org/jira/browse/SOLR-15673?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17428893#comment-17428893
]
David Smiley commented on SOLR-15673:
-------------------------------------
I haven't looked at the code but I suspect there is an assumption in the
implementation that the shard shape (specific shards w/ hash ranges) of the
cluster doesn't change, which isn't necessarily true. I suspect use of manual
sharing (aka implicit router) would be problematic as well. I suspect/worry
this is a hard problem to correct.
> Incremental backup attempt fails after a shard split operation has completed
> ----------------------------------------------------------------------------
>
> Key: SOLR-15673
> URL: https://issues.apache.org/jira/browse/SOLR-15673
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Components: Backup/Restore
> Affects Versions: 8.9
> Reporter: Sameer
> Priority: Major
> Attachments: StackTrace.odt
>
>
> I have been attempting to use the incremental backup API on Solr 8.9.0, but
> while testing in our product we would occasionally get into a state where all
> subsequent backup attempts would fail. After some triage we found that it was
> happening to any collection which had undergone a shard split operation. If
> we did a backup, completed a shard split operation, then attempted another
> backup, the second backup would fail with a FileNotFound exception relating
> to the backup id of the second backup as the error message.
> Steps to reproduce:
>
> * Create a new collection with no associated backups
> * Run a backup for this collection
> *
> /admin/collections?action=BACKUP&name=myBackupName&collection=myCollectionName&location=/path/to/my/shared/drive
> * Run a shard split operation
> * /admin/collections?action=SPLITSHARD&collection=name&shard=shardID
> * Attempt another backup
>
> Expected Outcome:
> * If this operation is being blocked intentionally, then I would expect an
> informative error message explaining why it failed. Otherwise I would expect
> the backup to complete successfully.
> Actual Outcome:
> * The backup operation fails with a NoSuchFileException.
> NOTE: In the below exception message the number in the file which isn’t found
> (in this case zk_backup_1) relates to the backup attempt which is currently
> being attempted.
> java.nio.file.NoSuchFileException:
> /path/to/my/shared/drive/reproCollectionBackup/reproCollection/zk_backup_1
> I tried a few different workaround attempts, but after going through these
> steps I wasn’t able to run another backup for the collection.
> Workaround attempt 1:
> * Use the API to delete the backup
> * Used the API to purge unused backup files
> * Restarted Solr
> * Attempted another backup
> * Encountered the same failure
> Workaround attempt 2:
> * Deleted all files in my Solr backup mount location
> * Restarted Solr
> * Attempted another backup
> * Encountered the same failure
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org
For additional commands, e-mail: issues-h...@solr.apache.org
