[
https://issues.apache.org/jira/browse/STORM-822?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15305342#comment-15305342
]
ASF GitHub Bot commented on STORM-822:
--------------------------------------
Github user jianbzhou commented on the pull request:
https://github.com/apache/storm/pull/1131#issuecomment-222308004
@hmcl, 1, work load is not distributed well is not because of the spout,
that is a kafka cluster setup issue and now is resolved 2, for the other two, I
dig into the log(sent to your via email) - seems everytime when a re-balance
happens, the spout seek to a bigger offset than the committed offset in this
partition, per my understanding, this will cause some message not be able to
consumed/emitted, so all the log show "Non continuous offset found"
user spout setting is: firstPollOffsetStrategy=UNCOMMITTED_LATEST,
pollTimeoutMs=2000, offsetCommitPeriodMs=10000, maxRetries=2147483647
I know firstPollOffsetStrategy cannot be EARLIEST or LATEST, but seems to
me UNCOMMITTED_LATEST should not cause this issue.
I asked user to try UNCOMMITTED_EARLIEST and now seems the issue does NOT
happen again as per the log, though it may happen later...
From the code perspective, i cannot understand why the weird behavior
happened, could you help?
Also, per our previous testing - we find once - a worker died and re
balance happened, we find one spout(not in the died worker) have some message
not acked or failed back. That also caused the "Non continuous offset found"
show many times in the log, which will cause no message will be committed to
kafka. The only solution will be restart the storm topology.
We emit message in this way - kafkaSpoutStreams.emit(collector, tuple,
msgId); Could you please help confirm - **storm would ensure all the messages
that emitted by the spout will be acked/failed back without exception?**
Because if this is not the case, the spout will not be able to find the
continuous offset to commit, then we must fix this issue urgently as we plan to
release the change early next month. Please help advise. thanks!
> Kafka Spout New Consumer API
> ----------------------------
>
> Key: STORM-822
> URL: https://issues.apache.org/jira/browse/STORM-822
> Project: Apache Storm
> Issue Type: Story
> Components: storm-kafka
> Reporter: Thomas Becker
> Assignee: Hugo Louro
> Fix For: 1.0.0, 2.0.0
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
