Can we not always reroute when update index setting?

[ViggoC]

Today I just want to update an index setting like refresh_interval, but it triggers a cluster rebalance.
Then I read the code and find the code here which always call reroute no matter what settings are updated. The comment says:

reroute in case things change that require it (like number of replicas)

SO can we only reroute when number of replicas (or other things that required it) are changed.

It doesn't look complicated. I can work on it. But I need to confirm that in addition to the number of replicas, are there any settings that need to trigger reroute?

[DaveCTurner]

There are many settings that require a reroute - allocation filters, tier preference, auto-expand replicas, disk watermarks, and so on, really anything that might require changes to the routing table. Also plugins can add their own allocation rules and index settings which may also require a reroute to be triggered on update. I think it would be safer to enumerate the settings that don't require a reroute since the consequences of missing a reroute are much worse than the consequences of doing an extra one.

Also note that there have been some optimisations around reroute in recent versions, so the value of this change needs evaluating against the code in master today. It may be that it's not worth pursuing.

Relates #77466, and #79866 would be a good improvement in this area too.

[elasticmachine]

Pinging @elastic/es-distributed (Team:Distributed)

[ViggoC]

I see. I think I can just disable cluster.routing.rebalance.enable to avoid unexpected rebalance.

[DaveCTurner]

Do you intend to work on any of the suggestions I made, or does your last message indicate that this can be closed?

[ViggoC]

I did not consider the need to support plugins before. I think this problem cannot be solved by modifying only one function. I'm closing this issue. Thanks for your time.