[
https://issues.apache.org/jira/browse/FLUME-2197?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14173062#comment-14173062
]
Johny Rufus commented on FLUME-2197:
------------------------------------
Given that Xmx is set to 6G, byteCapacity=0 is not correct, right ?
- I think thats true, if byte capacity is set to 0, then it defaults to
200GB, and if that much memory is not available, then it may lead to OOM
- (assuming that the max. no of events (capacity) has not yet been reached).
Shall we set byteCapacity to 50% of 6G or smaller ?
- setting this to 50% of 6G implies, you are only allocating 3G to be used by
the total set of events [header+body] in the memory channel
- if this is set, then this should be set to how much space you need for both
[headers+body] for the total number of events expected to be in the channel at
any point of time
- if this is not set, then it defaults to the heap size allocated to the jvm
- Also the byteCapacityBufferPercentage should be set close to the ratio of
header-size/[header-size+body-size]. Setting this to 50% means that every event
is expected to have equal header and body size
Hope this helps
> Memory Channel has GC issues
> ----------------------------
>
> Key: FLUME-2197
> URL: https://issues.apache.org/jira/browse/FLUME-2197
> Project: Flume
> Issue Type: Bug
> Reporter: Hari Shreedharan
> Assignee: Roshan Naik
> Attachments: HA_result.jpg, mem ch - mem alloc.png, spill ch - mem
> alloc.png
>
>
> Due to the fact that we use a LinkedBlockingDeque as the backing queue for
> the MemoryChannel, we end up hitting GC issues more often than we should.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
