[jira] [Commented] (OAK-8991) MarkSweepGarbageCollector: repeated warnings for files that don't exist

["Joel Kline (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OAK-8991?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17270482#comment-17270482 ] 

Joel Kline commented on OAK-8991:
---------------------------------

We currently have a repository when running MarkSweepGarbageCollector we see over 75,000 occurrences in our log files when this runs. It would be great if these references could be cleaned up when running. 

> MarkSweepGarbageCollector: repeated warnings for files that don't exist
> -----------------------------------------------------------------------
>
>                 Key: OAK-8991
>                 URL: https://issues.apache.org/jira/browse/OAK-8991
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>          Components: blob
>            Reporter: Thomas Mueller
>            Assignee: Amit Jain
>            Priority: Major
>
> When using the MarkSweepGarbageCollector (using for example a file data store), if the blob id file (from the BlobIdTracker) contains records that don't exist in the datastore, then a warning is logged when trying to remove the (unreferenced) file:
>  
> {noformat}
> *WARN* org.apache.jackrabbit.oak.plugins.blob.MarkSweepGarbageCollector Error occurred while deleting blob with id [...]
> org.apache.jackrabbit.core.data.DataStoreException: Record ... does not exist
> 	at org.apache.jackrabbit.core.data.AbstractDataStore.getRecord(AbstractDataStore.java:59) [org.apache.jackrabbit.jackrabbit-data:2.16.3]
> 	at org.apache.jackrabbit.oak.plugins.blob.datastore.OakFileDataStore.getRecordForId(OakFileDataStore.java:259) [org.apache.jackrabbit.oak-blob-plugins:1.8.9]
> 	at org.apache.jackrabbit.oak.plugins.blob.datastore.DataStoreBlobStore.getRecordForId(DataStoreBlobStore.java:520) [org.apache.jackrabbit.oak-blob-plugins:1.8.9]
> 	at org.apache.jackrabbit.oak.plugins.blob.datastore.DataStoreBlobStore.countDeleteChunks(DataStoreBlobStore.java:426) [org.apache.jackrabbit.oak-blob-plugins:1.8.9]
> 	at org.apache.jackrabbit.oak.plugins.blob.MarkSweepGarbageCollector$BlobCollectionType.sweepInternal(MarkSweepGarbageCollector.java:859) [org.apache.jackrabbit.oak-blob-plugins:1.8.9]
> 	at org.apache.jackrabbit.oak.plugins.blob.MarkSweepGarbageCollector.sweep(MarkSweepGarbageCollector.java:423) [org.apache.jackrabbit.oak-blob-plugins:1.8.9]
> 	at org.apache.jackrabbit.oak.plugins.blob.MarkSweepGarbageCollector.markAndSweep(MarkSweepGarbageCollector.java:287) [org.apache.jackrabbit.oak-blob-plugins:1.8.9]
> 	at org.apache.jackrabbit.oak.plugins.blob.MarkSweepGarbageCollector.collectGarbage(MarkSweepGarbageCollector.java:194) [org.apache.jackrabbit.oak-blob-plugins:1.8.9]
> {noformat}
> That means it tried to remove a file that doesn't exist.
> This indicates a problem in the process; for example, the blob id tracker file(s) was/were restored from an older backup. (Possibly there are other cases how this could occur).
> Now, the next time the garbage collection is run, the same files will try to be removed, and that again fails.
> It would be better if the files that don't exist are removed from the blob id tracker file, so that they are not tried to be removed later again and again.
> If the blob id tracker file(s) are incorrect, I think it would be better to delete and rebuild them, otherwise some of the unreferenced binaries will never be removed. Possibly a warning should be logged, with instructions on how to rebuild these files.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)