Working on rebasing the patch.

+1 for patch

Hi Ashish,

Thanks for the patch. I have some comments on this patch. Looks like this patch creates a ByteArrayOutputStream with initial size of 64K - similarly for the AdaptiveReceiveBufferSizePredictorFactory. I think we should start with a smaller buffer - probably at 2K and set the max to 64K.
So this can be AdaptiveReceiveBufferSizePredictorFactory (2048, 2048, 65536) and the ByteArrayOutputStream can be created using the current code itself. When the message is received, the current code already sets the maxSize to the event size.

I think the only change required to the original code is to use AdaptiveReceiveBufferSizePredictorFactory, but created with smaller buffer values. We can maybe ask the user for hints through config on average event size, but I doubt it is necessary.

Hi Hari,

Let me rework the patch and put it up for review again. I shall rebase the patch with latest code as well as incorporate the changes recommended.

Updated patch as per review comments. min size and initial size has been hardcoded to 2048. It's not configurable. We can alternatively make it configurable by asking from User. Please review and let me know if we need to do this.

The patch does not seem to apply (perhaps after yesterday's commit?).

Also, even now the SyslogUtils instance is being instantiated with maxSize as the size of the ByteArrayOutputStream - perhaps this should also be using DEFAULT_INITIAL_SIZE. BAOS does expand when needed.

Rebased
Updated as per review comment
The test class was changed, added new test case for larger layload

This looks good. The only issue is that UDP packets can't be fragmented, so even "large" packets may not reach the source at all based on the MTU of the netwrok being used. For example, I modified this patch's test and it was fine until I went up slightly about 7K (using the test in the patch itself).

+1. I am going to commit this. Thanks!

Commit e07a0a6883c84836e618d187c1381d47a26bfc71 in branch refs/heads/trunk from Hari Shreedharan
[ https://git-wip-us.apache.org/repos/asf?p=flume.git;h=e07a0a6 ]

FLUME-2130. Handle larger payloads via SyslogUDPSource

(Ashish Paliwal via Hari Shreedharan)

Commit 5de9a8af4d0f859bd88e65996e7543302fd9eb1f in branch refs/heads/flume-1.5 from Hari Shreedharan
[ https://git-wip-us.apache.org/repos/asf?p=flume.git;h=5de9a8a ]

FLUME-2130. Handle larger payloads via SyslogUDPSource

(Ashish Paliwal via Hari Shreedharan)

Committed! Thanks Ashish!

SUCCESS: Integrated in flume-trunk #547 (See https://builds.apache.org/job/flume-trunk/547/)
FLUME-2130. Handle larger payloads via SyslogUDPSource (hshreedharan: http://git-wip-us.apache.org/repos/asf/flume/repo?p=flume.git&a=commit&h=e07a0a6883c84836e618d187c1381d47a26bfc71)

• flume-ng-core/src/main/java/org/apache/flume/source/SyslogUDPSource.java
• flume-ng-core/src/test/java/org/apache/flume/source/TestSyslogUdpSource.java