5.13.3 is supposed to allow durable subscription rewrites, which should
allow garbage collection of old data files if the subscription messages are
all that's keeping the file alive. Having unconsumed real messages,
however, will still prevent garbage collection.
Do you have unconsumed real messa
Unfortunately, these errors was on production system, but I'll try to
reproduce it in test environment.
At this moment errors disappeared after I cleaned my KahaDB from old
messages (total size was about 8GB before cleaning because ActivemMQ 5.13.3
still doesn't properly cleanup journal files) an
You can also try setting this compiler flag: -XX:-OmitStackTraceInFastThrow
and seeing if that will show the inner stack trace as it might be an
optimization to hide it because log4j should usually log the inner stack
trace.
>From Oracle:
"The compiler in the server VM now provides correct stack b
This has been seen before (e.g.
http://activemq.2283324.n4.nabble.com/ClassCastException-while-subscribing-to-a-topic-tp4710870.html,
among others), but because we're not able to see the stack trace for the
inner ClassCastException, we don't know what's causing it.
The post I linked to is in the c
Maybe the kahadb files for that destination became corrupted somehow? Not
sure... is there any chance you can reliable reproduce this? Have you tried
setting ignore corrupted journal setting on kahadb to skip corrupted
sections when starting up the broker?
http://activemq.apache.org/kahadb.html
Matt-
Yes, the VM could be a root cause. I'm not a big fan of running
high-throughput messaging through VM's since this type of thing is
really hard to debug. The error you got sounded like it could be
related to I/O, so I definitely would try running it outside the VM.
I'm not intimately
Hi Matt,
Thanks for your reply.
We are just using standard queues. Nothing fancy. Unfortunately, I will not
be able to provide contents of the kahadb as it is rather large. Since we
fail under load, the KahaDB tends to be quite large (several GB).
This seems like a rare bug. Could it be related
Hi Matthew-
A negative seek offset sounds like a bug. I'd suggest creating a
ticket, and if you can attach the contents of the kahadb/ folder, that
would be helpful. Are you doing anything other than just simple
queues? Transactions, Durable Topics, Expiration, Message Groups, etc?
Matt P
