Hello all,
We are running a topology which reads data from kafka and interact with other
persistent stores.
incoming data into kafka is continuous and quite huge, but somtimes spout
throuput shown in storm UI becomes 0 which means entire topology hangs.
this event occurs rarely and we couldn't figure out what causes this, so it's
impossible for us to reproduce at this moment.
only weird thing we found so far is at that moment, load average and cpu
utilization got abnormally high.
we're to monitor worker process via JMX, is there any other methodology to
analyze this problem you would recommend?
Sincerely,
