Unassigned shards on elasticloud system indexes with 0 replica settings

Elastic Stack Elasticsearch
1

Using 1 node with number_of_replicas": "0", all beats indexes working properly, except system:

.monitoring-kibana
.watcher-history
.watches

It seem, that all system indexes have unassigned replicas, but all my settings seems correct.
As a side note, I created 2 zone environment, which I've shrinked to 1 zone.
POST _cluster/reroute?pretty
didn't help

Example of index template for .monitoring-kibana:

{
  "index": {
    "format": "7",
    "codec": "best_compression",
    "number_of_shards": "1",
    "auto_expand_replicas": "0-1",
    "number_of_replicas": "0"
  }
}

Yet all indexes are created with 1 replica
Other settings:
cluster.routing.rebalance.enable" : "all"

2

setting auto_expand_replicas to false will turn index green. What I don't understand is why this setting is not automatically adjusted for system indexes in managed cloud service?

3

How many shards do you have on the node? How much disk space have you got left?

4

74 shards, one per index + 20 yellow system ones.
plenty of space, 5-10% taken up on 2 nodes warm-hot.

5

Yellow status is not a problem as it just means that you have a replica shard configured which can not be allocated as you now only have one node.

6

I understand thank you, but I like things tidy, let's imagine a scenario where I have to prove the state of cluster is green and healthy.

If auto_expand_replicas defaults to false according to the documentation, then it seems it's an issue created by downsizing cluster from 2 zones to 1. If that's the case I guess it should be considered a bug and added to the elasticloud automation process.

7

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