[
https://issues.apache.org/jira/browse/OAK-2110?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14301095#comment-14301095
]
Marcel Reutegger commented on OAK-2110:
---------------------------------------
I tested the time to get documents marked with {{_deletedOnce}} with the
generic implementation and compared it with the mongo-specific one. The
mongo-specific version is roughly ten times faster. I think this is mostly
because it only has to execute the query once on MongoDB and can stream the
entire result to Oak vs. multiple queries to get the matching documents in
batches.
To solve this, we will probably have to change the signature of
DocumentStore.query() or introduce a new method.
> performance issues with VersionGarbageCollector
> -----------------------------------------------
>
> Key: OAK-2110
> URL: https://issues.apache.org/jira/browse/OAK-2110
> Project: Jackrabbit Oak
> Issue Type: Bug
> Components: mongomk
> Reporter: Julian Reschke
> Fix For: 1.1.7
>
>
> This one currently special-cases Mongo. For other persistences, it
> - fetches *all* documents
> - filters by SD_TYPE
> - filters by lastmod of versions
> - deletes what remains
> This is not only inefficient but also fails with OutOfMemory for any larger
> repo.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
