[
https://issues.apache.org/jira/browse/BOOKKEEPER-578?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13592536#comment-13592536
]
Ivan Kelly commented on BOOKKEEPER-578:
---------------------------------------
yes. It's a documentation issue. I admit that before this issue cropped up, I
assumed that direct memory was counted on the heap stats. We have 2 options.
Document this better (which might involve updating the scripts to warn if mx is
set too high), or switch to using non direct buffers. Direct buffers are slower
to allocate, but avoid the read. But maybe the read isn't an issue.
> LedgerCacheImpl is reserving 1/3 of Heap size but allocates NonHeap memory
> --------------------------------------------------------------------------
>
> Key: BOOKKEEPER-578
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-578
> Project: Bookkeeper
> Issue Type: Bug
> Components: bookkeeper-server
> Reporter: Matteo Merli
> Priority: Minor
>
> By default the page limit parameter is set to -1, which means to assign 1/3
> of Heap space to the LedgerCache. Each LedgerEntryPage is then allocating the
> memory outside the heap (ByteBuffer.allocateDirect()).
> This makes BK to use more memory than the -XmxNN configured setting. Is there
> any particular reason for the LedgerEntryPage buffer to be allocated outside
> the java heap? Could that be changed?
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
