jackrabbit-oak-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Julian Reschke (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (OAK-6303) Cache in CachingBlobStore might grow beyond configured limit
Date Fri, 02 Jun 2017 14:18:04 GMT

    [ https://issues.apache.org/jira/browse/OAK-6303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16034766#comment-16034766
] 

Julian Reschke commented on OAK-6303:
-------------------------------------

One thing to do is to make sure that the number of segments times the blocksize isn't bigger
than the maximal cache size. Right now we have 16 segments and 2 MB blocksize, so once every
segment holds one block we already exceed the max memory size.

To fix this, we could use something like
{code}
                  segmentCount((int)(cacheSize / getBlockSize())).
{code}
in the constructor of {{CachingBlobStore}}.

The other thing to look at is how the individual segments behave when trying to evict entries
and the number of entries is small (which will be the case in the above test).

> Cache in CachingBlobStore might grow beyond configured limit
> ------------------------------------------------------------
>
>                 Key: OAK-6303
>                 URL: https://issues.apache.org/jira/browse/OAK-6303
>             Project: Jackrabbit Oak
>          Issue Type: Bug
>          Components: blob, core
>            Reporter: Julian Reschke
>            Assignee: Thomas Mueller
>         Attachments: OAK-6303-test.diff
>
>
> It appears that depending on actual cache entry sizes, the {{CacheLIRS}} might grow beyond
the configured limit.
> For {{RDBBlobStore}}, the limit is currently configured to 16MB, yet storing random 2M
entries appears to fill the cache with 64MB of data (according to it's own stats).
> The attached test case reproduces this.
> (it seems this is caused by the fact that each of the 16 segments of the cache can hold
2 entries, no matter how big they are...)



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message