I am trying to run in K8s with the latest updates ...
# all-things-deployment
s
I am trying to run in K8s with the latest updates and I am seeing an odd behavior. I did helm uninstall and then helm install and now MCE is trying to process old messages that do not exist anymore. I am confused as to where the MCE is getting these old messages since the topics are empty.
m
@early-lamp-41924: thoughts?
e
Did you uninstall prerequisites as well? Kafka uses a persistent store, and even if you reinstall it it will pick it up again
s
We are using Kafka outside of K8s. I lowered the message timeout and then increased it after a few minutes.
That should have cleared it. We also have monitoring on the topics and can see that there is no activity.
It looks like the offset was set back to zero but the messages still exist. Let me confirm.
That was it. Sorry for the bother.
m
That's great to hear. @sticky-television-18623: any thoughts on what a more improved debugging experience might look like?
s
Off the top of mind I would say more detailed logging on exceptions. For example, right now I looking at an MCE error:
Copy code
19:25:30.988 [mce-consumer-job-client-0-C-1] ERROR c.l.m.k.MetadataChangeEventsProcessor - MCE Processor Error
java.lang.ClassCastException: com.linkedin.r2.RemoteInvocationException cannot be cast to com.linkedin.restli.client.RestLiResponseException
This does not tell me much about what the cause of the exception might be.
m
thanks! we'll take a look here