Hi Tim, Overall looks good to me. we recommend adding supervisord ( a watchdog process) to monitor nimbus and supervisors and restart them incase if they fail but since you already upstart which does the same job. Thanks, Harsha
On June 18, 2015 at 1:15:16 PM, Tim Molter ([email protected]) wrote: Hi Storm gurus, I recently took the plunge into running a topology on a cluster with 4 hosts. I was hoping someone could critique my installation set up and deployment which I wrote up here: http://knowm.org/how-to-install-a-distributed-apache-storm-cluster/ In particular, is setting up zookeeper, nimbus, storm-ui, and supervisor as an Upstart service a good way to run things "under supervision" as described in the Storm docs? The docs say "Additionally, as the cluster is "self healing" the failed server once restarted will automatically rejoin the ensemble w/o any manual interaction." Is my respawning Upstart service an appropriate approach to this? Any other comments on my setup? I appreciate any feedback very much. thanks in advance.
