Would it make sense to increase the 'stale_master_timeout' setting to
something like 5 minutes? What would be the issues to consider with a
large cluster (say 32 Graylog Server nodes) having this set at 5 minutes
(instead of 2000ms)?
My understanding is that the master is only needed to run s
Hello Jochen,
Thanks for the response and paraphrase explanation, that helped me make
more sense of what was going on. I took another look at my NTP
configuration and as it turns out, the system clock wasn't syncing as it
should have been. I fixed that, and the warnings from graylog stopped.
Hi Ariel,
just for reference, I'll paraphrase the explanation from IRC:
Each Graylog node "registers" itself (node id, URI to the Graylog REST API,
> timestamp of the last heartbeat) in MongoDB (see the nodes collection).
> The timeout/cleanup interval is quite aggressive (2s, see
> stale_mast
