Upgraded from 2.4.6 to 5.3.2 - now fails to form the cluster

Did a cluster shutdown (all nodes) and then upgraded node by node from 2.4.6 to 5.3.2 and started cluster again node by node, but they all just continually complains like this:

[2017-10-26T23:21:48,072][WARN ][o.e.d.z.ZenDiscovery ] [d1r2n9] not enough master nodes discovered during pinging (found [[Candidate{node={d1r2n9}{jJ3HFWbhSfudgfaK4w-y8A}{yCVvctQ3TR6ye1k9txj6cg}{<ip>}{<ip>:9300}{rack=OPA3.4.16}, clusterStateVersion=-1}]], but needed [8]), pinging again

even though I've restart all 14 nodes again.

In 2.4.6 we used:

discovery.zen.ping.multicast.enabled: false
discovery.zen.ping.unicast.hosts: [<list of nodes>]

but had to remove this to start 5.3.2 nodes, believe I've read somewhere than 5.x now only uses unicast for cluster communication anyway.

Hints appreciated to rejoin cluster, TIA!

Solved: still had to defined the ping unicast host list :slight_smile:

Not sure if the multicast.enabled setting is support, but unicast hosts is
definitely important. With multicast, nodes can dynamically find each
other, but with unicast, an explicit list of hosts is needed. Perhaps
increase the timeout so that you can start the minimum amount of nodes in

Isn't support according to the migration helper plugin, hence removed it, but mistakenly also removed host list whih is needed for unicast

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.