Github user srdo commented on the issue:
https://github.com/apache/storm/pull/1679
@apiwoni Okay, just wanted to know because 1.0.x doesn't handle topic
compaction very well, and I thought it might be related.
Your log looks a lot like this issue. I'm not sure why you expect the poll
to return records after seekToEnd? seekToEnd seeks to the end of the log in
Kafka, and it doesn't actually do the seek until poll or position is called on
the consumer. The result is that the consumer's position is at the last offset
in the log, which means it won't emit anything until new messages are added to
Kafka.
Due to this issue it will probably cause the spout to stop emitting
entirely, since the failed offsets keep triggering the seekToEnd, which makes
the spout skip past all new messages every time it polls.
I verified this behavior with a small example application with a single
consumer that just calls seekToEnd and poll in a loop. It never emits any
messages.
You should upgrade the spout. Whether you want to upgrade the rest of Storm
is up to you, but be aware of https://issues.apache.org/jira/browse/STORM-2682
before deciding to jump to 1.1.1.
---
