Github user kevinconaway commented on the issue:
https://github.com/apache/storm/pull/2182
Thanks for your patience on this. I did some debugging on Travis on my
fork and found 2 issues
- The behavior of binding to _localhost_ seems somewhat inconsistent for
Zookeeper and Kafka. I noticed a bunch of ZK connection issues trying to
connect to the hostname of the travis machine that wouldn't resolve. Using
_127.0.0.1_ fixes this
- More importantly, there were timing issues related to creating the topic
and the partition leaders being assigned.
`DynamicBrokersReader#getNumPartitions` will fail if the topic doesn't yet have
leaders assigned to its partitions and this was the root of the NPE.
I fixed the second issue by waiting until the topic is fully ready in kafka
(exists in zookeeper and all partitions have leaders) before attempting to use
it. I also adjusted the logic around emitting messages from the partition
manager to allow for timing delays on the broker end as this was another cause
of spurious failures.
I've also adjusted the _integration-test_ parent pom version to match the
current one in the branch. It looks like this was missed when the version was
bumped to _1.2.0-SNAPSHOT_
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---