Hi,
My first post here. Right now I'm trying to use Heka as a logstash
replacement (upgrade :) in a kafka-ELK/Cassandra message processing
system.
It seems like if heka doesn't shut down cleanly then the result can be
zero length partition
offset files. At least that's my assumption of how the files got
created.
The real issue arises when heka (re)starts and the zero length file
results in a readCheckpoint error.
I'm wondering if this is the right behavior as right now I don't
restart the heka service I
use a three step; stop, delete zero length offset files and restart,
process.
Would it be better to treat zero length or corrupt file as a warning
and proceed as if the file never existed?
Thanks,
Ade
_______________________________________________
Heka mailing list
[email protected]
https://mail.mozilla.org/listinfo/heka
