Storm finds leader Nimbus via connecting to known Nimbus and query, so it's
safe unless all of Nimbus in nimbus.seeds are dead.
But, AFAIK, runtime update of Nimbus seeds (via discovery) doesn't happen.
(Anyone please correct me if I'm wrong.)
Might be better to have it for H/A (run discovery for given interval, or
whenever connection to Nimbus), but not sure we would also want to also
update the config file.
Jungtaek Lim (HeartSaVioR)
2016년 10월 13일 (목) 오후 7:57, Satish Duggana <satish.dugg...@gmail.com>님이 작성:
> Which version of Storm are you using? If you are using 1.x then you can use
> nimbus HA by configuring respective nimbus.hosts.
> On Thu, Oct 13, 2016 at 11:56 AM, S G <sg.online.em...@gmail.com> wrote:
> > Hi,
> > We are setting up a storm cluster where a nimbus node would be
> > automatically added to the nimbus cluster if any nimbus machine fails.
> > In such a scenario, the new nimbus node will have a different
> > address.
> > If that happens, is it absolutely mandatory for all the
> > workers/supervisors/remaining-nimbuses to have their storm.yaml files
> > updated immediately?
> > Or they can get to know about the new nimbus node from the existing
> > nodes?
> > (We will update the storm.yaml for any node that is restarting, but it is
> > somewhat difficult to trigger the updation of storm.yaml whenever a new
> > nimbus node joins the nimbus cluster).
> > Some help on the same would be appreciated.
> > Thanks,
> > SG