A high-disk watermark protects an Elasticsearch node when storage pressure becomes unsafe. At the default 90% threshold, Elasticsearch alerts and works to relocate shards to nodes with adequate disk space. For example, a node with 1 TB allocated and 850 GB consumed approaches the configured protection limits. section 15.3.3
Listing 15.15 Setting the high-disk watermark on the cluster
PUT _cluster/settings
{
"transient": {
"cluster.routing.allocation.disk.watermark.high":"0.85"
}
}
Change the high-disk watermark dynamically through the cluster settings API; the example expresses an 85% threshold as 0.85. section 15.3.3
Relate the high-disk watermark to the surrounding disk-protection thresholds:
- Use the low-disk watermark to manage shard allocation and signal emerging storage pressure. section 15.3.3 section 15.6
- Expect shard relocation when the high-disk watermark is crossed. section 15.3.3
- Treat the flood-stage watermark as the emergency threshold: affected shards become read-only and indexing stops. section 15.3.3 section 15.6
