Rolling upgrade 6.8.1

I have a 6 node cluster where all nodes are master eligible and data nodes
I have 2 degrees of Replication across the cluster
During rolling upgrade from 6.2.4 to 6.8.1 is it ever possible for the cluster to be green in the upgrade process?
if the 1st node is upgraded, and all other nodes are on older versions, the cluster goes to yellow state because during the upgrade process, once shards get allocated to nodes with newer versions, these same shards cannot be reallocated to nodes with lower versions of elastic (normal behavior)

I just want to understand if this means through the rolling upgrade process my cluster will always be yellow until all nodes are upgraded and have same versions , is this correct please ?

Yes, if a new index is created, it will be created on nodes with the highest Lucene and it's replica's must be on a matching Lucene version, ,so at this time, they CAN'T be created.

How much time passed? A rolling upgrade probably needs to be completed within the (daily, whatever) cycle of new index allocation.

As you upgrade more nodes, the replicas should be created.

1 Like

Thank you , much appreciated for the answer :slight_smile:

As for the time passed by? I would say alot of time would have passed by in my case, and would exceed the cycle of new index allocation.
What would this mean then for me please ?

In cases where i know upgrades will last a long time, is it recommended for me to stop index creation through out the rolling upgrade process then

What is a reason to not proceed to the next nodes as soon as the first is upgraded?

Delaying gives you the choice, don't index or index but don't have replication. Only your environment can make that choice.

1 Like

thank you again.
I will avoid delays during the upgrade process between the nodes then.

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