[
https://issues.apache.org/jira/browse/METRON-538?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15677761#comment-15677761
]
ASF GitHub Bot commented on METRON-538:
---------------------------------------
Github user ottobackwards commented on the issue:
https://github.com/apache/incubator-metron/pull/344
@dlyle65535
OK, I think I know what is going on.
We kind of have a mess wrt the zookeeper stuff in the components. I had
originally taken a stab at untangling it, but had problems and reverted back
and scaled back my attempts. Now with my other PR for Yarn and
ZKServerComponents - I'm going to try again.
The issue is that the KafkaWithZK component creates a zkserver, but it
immediately goes out of scope.
I believe that causes the AdminUtils to fail looking up the brokers.
Now that I have a good ZKComponent ( from curator as opposed to the Kafka
testing server one ) I'm going to try again
> Ensure proper shutdown, exceptions of integration test components
> -----------------------------------------------------------------
>
> Key: METRON-538
> URL: https://issues.apache.org/jira/browse/METRON-538
> Project: Metron
> Issue Type: Bug
> Affects Versions: 0.2.2BETA
> Reporter: Otto Fowler
> Assignee: Otto Fowler
> Priority: Critical
>
> The integration test components use to test with kafka, in mem cluster and
> zookeeper can be left in a failed state, where they are not shut down and
> their resources released. This can lead to cascading failures in the testing
> environment and log pollution.
> The components and their use should be refactored such that they are always
> shut down properly, and log appropriately.
> Some examples will be added.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
