0.90.7 - split brain right out of the box


(revdev-2) #1

Hi everyone,

set up a new 0.90.7 cluster with 5 machines. here are the relevant parts of
the config:

cluster.routing.allocation.node_initial_primaries_recoveries = 4
cluster.routing.allocation.node_concurrent_recoveries = 2
discovery.zen.minimum_master_nodes = 3
gateway.recover_after_nodes = 3
gateway.expected_nodes = 5

the cluster was spun up using chef and the 5 processes came online about
the same time (within 10s of each other) and immediately formed into a
split brain where 4 nodes formed a cluster and one member decided to elect
itself a master in a 1-member cluster. restarting that node caused it to
join properly. my question is, how would it be possible for this scenario
to occur given the settings above? does an es cluster require rolling
start/restart?

thanks!
-wayne

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


(system) #2