I have a similar problem.
I tried both settings but the spout is still running OOM, because the
following bolts cannot consume the tuples due to slow I/O (in my case
network).
Since I'm not using anchoring, might this cause the problem? Also, my spout
shows 0 acked tuples, is this normal using the non-anchored option?
Btw, where is the best place to set those configs? I was setting the max
spout pending thing both in the topology and in the config of the master,
nodes and my local storm.yaml. Which one is used when deploying topologies?
(I cannot use storm-deploy)
Thanks,
Michael
Am Dienstag, 18. Dezember 2012 10:57:33 UTC-8 schrieb Vinh:
Have you tried setting "topology.max.spout.pending" to limit the amount of
pending tuples? Else, spouts will keep pushing tuples to the bolts,
regardless of result.
On Dec 18, 2012, at 10:50 AM, Davide Simoncelli wrote:
Hi Nathan,
I have a problem: few node that run a topology crash because Storm goes
in out of memory.
I tried to set the High water mark limit of ZMQ queue in storm.yaml
(zmq.hwm) so that when the queue limit is reached the receiver or sender
should block. Unfortunately this solution doesn't work. Do you know why?
Thank you
- Daviode
