Going from 6.8.15 to 6.8.18 should be no issue correct

Good Day,

I have a cluster that has ES 6.8.15 install and see that 6.8.18 is there to upgrade to. This should be a straight forward upgrade correct. Everything is currently working in our cluster, but I wanted to make sure there were no gotchas that I’m overlooking before updating. Eventually, we will move to 7.x but that another day update.

Also, just for my knowledge. Let’s say I have my current nodes at 6.8.15 and add a new node that’s at 6.8.18… Will that break the cluster cause one node is at a higher version?

Also, to upgrade, I should be able to simply upgrade the nodes and reboot the cluster correct (I don’t have to do a rolling upgrade, I can bring it down temporarily). Is my thinking correct?

Thanks,

As long as you are following the guidance in Upgrade Elasticsearch | Elasticsearch Guide [7.14] | Elastic, you'll be fine.

Thanks,

Is it possible to have one node be at a point version higher than the other nodes? (Example: one node at 6.8.18 while the other are at 6.8.15).

Also, the upgrade guide doesn’t really address 6.8.15 to .18. But I guess I could stop the service on each node and upgrade the installs on the server to get .18 and turn on the service again.

One last question,
Should I do the upgrade in any particular order if you have a cluster of specialized nodes?

Thanks

Yep, that's the concept behind a rolling upgrade. You just cannot go backwards once you have upgraded that node.

Generally you'd do;

  1. Any master only nodes, leaving the active master till last
  2. Data nodes
  3. Client/ingest or other node roles

If you have a homogenous cluster (ie all nodes play all roles), just do every node other than the active master, and do it last.

All nodes in the cluster should always use exactly the same version.