[
https://issues.apache.org/jira/browse/SAMZA-458?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chris Riccomini updated SAMZA-458:
----------------------------------
Attachment: SAMZA-458.2.patch
Attaching a slightly tweaked patch that changes SamzaException to include the
exception as a second param, rather than a {{%s}} field in the message.
[~navina], I had a question about this line as well:
{code}
sendFailed.set(false)
{code}
Is it safe to do this? I'm wondering if it's possible for sendFaield to be set
to true, and then send is called again, where it's reset to false. I *think*
that this shouldn't happen, since the Callback is only invoked from the main
thread, not from Kafka's sender thread (since we have max retries set to
Int.MAX). Correct?
Second question: if we ever get a RetriableException thrown, don't we go into
an infinite loop? In this case, sendFailed is true, and we loop back around,
where the first line in the loop is:
{code}
if(sendFailed.get()) {
throw exceptionThrown.get()
}
{code}
I don't see how we escape this. I think we don't see this now because we never
get RetriableExceptions because we have retries sent to Int.MAX.
> Close in KafkaSystemProducer should flush all source buffers
> ------------------------------------------------------------
>
> Key: SAMZA-458
> URL: https://issues.apache.org/jira/browse/SAMZA-458
> Project: Samza
> Issue Type: Bug
> Components: kafka
> Reporter: Chris Riccomini
> Assignee: Yan Fang
> Labels: newbie
> Fix For: 0.9.0
>
> Attachments: SAMZA-458.1.patch, SAMZA-458.2.patch, SAMZA-458.patch
>
>
> I noticed that calling KafkaSystemProducer.stop does not flush any
> outstanding messages in the sourceBuffers array. Calling close should flush
> all buffers. Without this, shutting down can cause messages to be dropped.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
