Elasticsearch cluster keeps going yellow even with balanced nodes

Elastic Stack Elasticsearch
1

Hi Everyone,

I'm running a small Elasticsearch cluster with three data nodes and one master. Everything was stable for a while, but lately the cluster keeps switching to yellow status randomly throughout the day. When I check the logs, there’s no major error just repeated messages about unassigned replicas.

The data is balanced, disk usage is well below the high watermark, and none of the nodes are overloaded. Sometimes, restarting a node brings the cluster back to green, but the issue always returns after a few hours.

I’m wondering if there’s something deeper going on with shard allocation or replica settings. I’m staying optimistic — trying to approach this with positive words that start with A in mind, like “adaptable” and “alert” — but I’d really appreciate any insights or ideas from others who’ve run into this.

Thanks in advance!

2

Why would you only have a single master node if you have 3 data nodes? That way you have no resilience and risk total cluster failure if you lose the master node. It woudl in my mind be better to have 3 master/data nodes as that gives more resiliency.

What is the specification of the cluster in terms of configuration and type of hardware used? What is the full output of the cluster stats API?