ECK Operator not resetting shard_allocation setting to "all"


I've always had the the problem that the ECK operator does not reset the shard_allocation cluster setting after restarting a node. The result is that after a rolling restart over night, I end up with hundreds of unallocated replica shards and no redundancy, because transient.cluster.routing.allocation.enable is set to primaries. This will also halt the rolling reboot at some point when no more nodes can be restarted without making the cluster red.

This basically means that for each cluster reboot, I have to stand by all the time and issue a

PUT _cluster/settings
  "transient": {
    "cluster.routing.allocation.enable": null

after each node, only so the operator can reset it to primaries as soon as the next node is restarted.

Is this a known issue? For me this issue has been around since the beginning and no update has solved it so far.

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