[
https://issues.apache.org/jira/browse/OAK-9765?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17533263#comment-17533263
]
Piercarlo Slavazza commented on OAK-9765:
-----------------------------------------
I would like to add comments derived from some more tests that I did:
* I had this (naive?) idea of retrieving the Blobs in the NodeStore by
searching via a SQL2 query that searches for nodes that have {{{}jcr:data{}}}:
this way, the used Blobs are properly marked, and eventually swept by the GC
* I also found another issue: the {{MarkSweepGarbageCollector}} raises an
error when there are no Blobs marked; this seems wrong to me… maybe I miss some
piece of the while thing?
I added these patches to the Github repo linked above.
If someone could confirm the fitness of my findings/solution, I would be happy
to write a patch and a PR (in case, I would need some help in understanding the
best way to write unit tests).
> Garbage Collection does not remove blobs file from the file system
> ------------------------------------------------------------------
>
> Key: OAK-9765
> URL: https://issues.apache.org/jira/browse/OAK-9765
> Project: Jackrabbit Oak
> Issue Type: Bug
> Affects Versions: 1.42.0
> Reporter: Piercarlo Slavazza
> Priority: Blocker
>
> Using a NodeStore backed by a FileStore, with a blob store of type
> FileBlobStore:
> # (having configured GC with estimation {_}disabled{_})
> # a file is added as a blob
> # then the node where the blob is references is _removed_
> # then the GC is run
> # expected behaviour: the node is no more accessible, _and_ no chunk of the
> blob is present on the file system
> # actual behaviour: the node is no more accessible BUT all the chunks are
> still present on the file system
> Steps to reproduce: execute the (really tiny) main in
> [https://github.com/PiercarloSlavazza/oak-garbage-collection-test/]
> (instructions in the readme)
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
