[
https://issues.apache.org/jira/browse/KAFKA-10843?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17254507#comment-17254507
]
Maria Thomas commented on KAFKA-10843:
--------------------------------------
So from the Logging tools associated with the Kafka clusters, I see that two of
my topic's partitions are associated with a single instance and the other one
with the second instance. The third instance has no partitions assigned to it.
I can also see the requests being distributed between the two instances and
processed successfully.
However, while there should be two state stores at least to be in synch with
the above observation, allMetadata() returns only one instance(which had two
partitions assigned to it) with the required state store while the other two
instances(including the one with the third partition) are empty.
> Kafka Streams metadataForKey method returns null but allMetadata has the
> details
> --------------------------------------------------------------------------------
>
> Key: KAFKA-10843
> URL: https://issues.apache.org/jira/browse/KAFKA-10843
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Affects Versions: 2.5.1
> Reporter: Maria Thomas
> Priority: Major
>
> Our application runs on multiple instances and to enable us to use get the
> key information from the state store we use "metadataForKey" method to
> retrieve the StreamMetadata and using the hostname do an RPC call to the host
> to get the value associated with the key.
> This call was working fine in our DEV and TEST environments, however, it is
> failing one our production clusters from the start. On further debugging, I
> noticed the allMetadata() method was returning the state stores with the host
> details as expected. However, it would not be feasible to go through each
> store explicitly to get the key details.
> To note, the cluster I am using is a stretch cluster.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
