User Guide

Managing an Elasticsearch Cluster with a Full Disk

If a cluster node’s disk begins to fill up (defined as going below a low disk watermark, with the value for cluster.routing.allocation.disk.watermark.low set by default to 85%), then replicas will no longer be assigned to that node by setting the flag read_only_allow_delete. This may cause unexpected behavior because new indices and shards may not be allocated or replicated correctly.

Once more space is available for the Elasticsearch data (e.g., you have increased the available space for the logical volume on the cluster), you should reset the value to normal to allow the node to return to normal function by using the following command:

# /usr/share/neteye/elasticsearch/scripts/ 'https://elasticsearch.neteyelocal:9200/*/_settings' -X PUT -H 'Content-Type: application/json' -d'
       "index.blocks.read_only_allow_delete": null
     ' | jq .