I just noticed my ES 6.2.4 cluster not recovering from yellow after a recent node failure. Some shards would remain unassigned.
Upon closer inspection, the affected indices had index.allocation.max_retries set to 1, which was apparently a side effect of a Shrink operation I ran a week ago.
Why is this parameter being set by Shrink?
Resetting it to null made the cluster instantly recover the shards back to green.
Is there any harm in doing so?
Are there any other hidden settings that Shrink leaves behind that I should know about?