Rolling Upgrade from 6.8.0 to 7.3.1 Failed to Join the Cluster

Elastic Stack Elasticsearch
1

Hello all,
I am currently having an Elasticsearch 6.8 Cluster with one master and one slave node.
I need to upgrade this Elasticsearch to current version 7.3.1.
I am following the Rolling Upgrade Steps given in https://www.elastic.co/guide/en/elasticsearch/reference/current/rolling-upgrades.html

But after upgrading the Master node, unable to Join it with the Slave node. Getting the following error:
[2019-09-09T11:16:37,868][INFO ][o.e.d.z.ZenDiscovery ] [node-2] failed to send join request to master [{node-1}{rhnmAFjARd-OGXRy_eRiHQ}{GRwn2uehQ7awr6QujUgBYw}{192.168.2.10}{192.168.2.10:9300}{ml.machine_memory=8438468608, ml.max_open_jobs=20, xpack.installed=true}], reason [RemoteTransportException[[node-1][192.168.2.10:9300][internal:discovery/zen/join]]; nested: IllegalStateException[index [.kibana_1/GXKbMQisTte_Zswjiks_xA] version not supported: 7.3.1 the node version is: 6.8.0]; ]

The Rolling Upgrade says that we don't need to Restart the Full Cluster but here I am getting such issue. Can anyone tell me if I am following any wrong approach?

2

You only have one master eligible node so you cannot perform a true rolling upgrade: when you shut the master-eligible node down you are effectively fully restarting the cluster. You should follow the instructions for a full cluster restart upgrade instead.

I think the docs could be clearer on this point.

3

Elasticsearch is a clustered syystem and does not have master and slave nodes. In order to have a highly available system you need at least 3 master-eligible nodes. These can also hold data in order to keep the cluster size down.

4

Thanks for your quick reply.
But with the same Master- Slave setup I have successfully upgraded the Cluster from 6.2.4 to 6.8 following Rolling Upgrade instructions given at https://www.elastic.co/guide/en/elasticsearch/reference/6.2/rolling-upgrades.html

5

Yes I think I'd expect that to work from 6.2 to 6.8, although it wasn't a true rolling upgrade since the cluster would have stopped when you shut the master node down. Elasticsearch can be lenient when upgrading to a new minor version, but ithas to be stricter when moving to a new major version.

1 Like
6

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