storm-user mailing list archives

Sometimes when we run a three node topology, if a worker fails and comes back up, the entire
topology will become sluggish, and messages will constantly be marked as failed. After changing
the logging, we can determine that the topology is actually fully processing messages, however
they are never being passed back to the acker to be acked. I've done searches to try and find
solutions (other than don't let the worker fail) to fix the issue, but haven't found anything
yet.