Hi,
we are using KafkaSpout from Wurstmiester for 3 topologies. For one of the
topologies, the spout has a very high failure rate.
For the first two topologies, we see that the kafka spout has a "complete
latency" of under 100 mse (0.9 msec and 58 msec). The third topology has
the complete latency in Storm UI at 11 sec. The default value for
TOPOLOGY_MESSAGE_TIMEOUT_SECS
is at 10 sec. I guess this is the reason for failure.

Question: What are the reasons for a spout having such high "Complete
Latency"? Does it indicate that the bolt which gets tuples from this spout
has a "bottleneck" and hence unable to accept tuples and hence failures? or
could there be other reasons?

Please help!

Thanks,
Kashyap

Reply via email to