Don't ignore allocation settings with cluster reroute

[clintongormley]

The cluster reroute command ignores the allocation deciders for index/cluster allocation settings, which makes it hard to debug why shards are not being assigned (see #18294 (comment))

[dakrone]

I believe it should be an option to disable the allocation deciders when running reroute commands. I've disabled allocation and then manually allocated shards to be where I want them to be before in order to reproduce a specific cluster allocation

[clintongormley]

While that may be useful from a debugging perspective, it sounds very trappy for users The least we should do is report them as ignored in the explain output

Update: I was missing the context. I'm +1 on providing a flag to allow disabled allocation to be overridden with the cluster reroute command

[s1monw]

Update: I was missing the context. I'm +1 on providing a flag to allow disabled allocation to be overridden with the cluster reroute command

+1 for now but in general we should really try to get away from providing this API alltogether

[clintongormley]

The flag should also allow overriding of the index.allocation.max_retry counter, which is being added in #18467

[bleskes]

heads up to whoever implements this - some allocation deciders should never be overridden - for example ReplicaAfterPrimaryActiveAllocationDecider and SameShardAllocationDecider. These are so essential to how the code works, that I wonder if we should give them a special place/hard code them.

[elasticmachine]

Pinging @elastic/es-distributed

[DaveCTurner]

which makes it hard to debug why shards are not being assigned

We think this is now covered sufficiently well by the allocation explain API.