[
https://issues.apache.org/jira/browse/FLUME-742?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tom White updated FLUME-742:
----------------------------
Attachment: FLUME-742.patch
This patch fixes the problem by changing AvroEventSink so it doesn't null out
the underlying transport (per offline discussion with Jon).
I also updated the test so the secondary failure is avoided by weakening the
assertion to check only that the number of bytes sent over the wire is at least
the message size (it will be more due to headers).
I noticed that TestThriftSinks was failing too, but after fixing TestAvroSinks
it no longer failed, even after I looped the test for a few minutes.
It looks like this failure went unnoticed since Maven doesn't mark the test as
failed after killing it:
{noformat}
Running com.cloudera.flume.handlers.avro.TestAvroSinks
Process 1313433960885 is killed.
{noformat}
It would be good to fix that too so that tests that time out don't go unnoticed.
> TestAvroSinks fails
> -------------------
>
> Key: FLUME-742
> URL: https://issues.apache.org/jira/browse/FLUME-742
> Project: Flume
> Issue Type: Bug
> Reporter: Tom White
> Attachments: FLUME-742.patch, FLUME-742.patch
>
>
> The following appears in the log
> {noformat}
> 2011-08-12 21:21:48,784 [Thread-13] INFO avro.AvroEventSink: AvroEventSink
> open on port 35854
> 2011-08-12 21:21:48,863 [Thread-13] INFO debug.MemorySinkSource: Closing mem
> source sink that has 25 elements
> 2011-08-12 21:21:48,863 [Thread-13] INFO avro.AvroEventSink: AvroEventSink on
> port 35854 closed
> Exception in thread "Thread-13" java.lang.NullPointerException
> at
> com.cloudera.flume.handlers.avro.AvroEventSink.getSentBytes(AvroEventSink.java:116)
> at
> com.cloudera.flume.handlers.avro.TestAvroSinks$3.run(TestAvroSinks.java:196)
> {noformat}
> Then the test (testManyThreadsAvroSend) times out.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
