Github user jfenc91 commented on the issue:
https://github.com/apache/storm/pull/1679
@srdo @hmcl I tracked down one source of the "double acking"
It looks like I am being switched to reading from an out of sync replica.
Looking at my offsets:
GROUP TOPIC PARTITION
CURRENT-OFFSET LOG-END-OFFSET LAG OWNER
bla_5_2 bla_5 102 237919 50979
-186940 consumer-7_/123.456.789.101
When examining with my other partition logs there is no way that this
log-end-offset should be at 50k. It should be around 200k-300k. When the
connector sends the request for offset 237920 it is out of range and the resets
the offset to whatever the start offset is:
https://github.com/apache/kafka/blob/0.10.0.1/clients/src/main/java/org/apache/kafka/clients/consumer/internals/Fetcher.java#L580-L589
. For me this happened to be uncomitted_earliest so I got messages starting at
position 1.
This is a really unfortunate situation. We should probably make an effort
to handle it better so data loss / disruption is minimized. I kind of wonder if
this ever happened with the kafka 8.
---
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.
---
