When doing a rolling restart we recommend disabling shard allocation to
avoid unnecessary recoveries. However, this advise is unnecessary or
even harmful when restarting nodes that do not carry any data like a
pure ML node.
The current documentation recommends disabling allocations of all shards. This prevents shards of
new indices from being allocated as well. That means that during a rolling upgrade, operations like
roll over and index shrinking will operate correctly. This, in turn, can cause issues for data ingestion
and ILM.
