High memory-to-disk ratio on frozen nodes
Frozen-tier data nodes show a higher memory-to-disk ratio than the recommended range for that role. Frozen search is largely I/O-bound, so surplus heap rarely improves query speed enough to justify the size.
For a complete list of insights, refer to AutoOps insights.
| Field | Value |
|---|---|
| Component | Elasticsearch |
| Severity | Low |
| Scope | Node |
| Domains | memory, resource-utilization |
You can customize these settings to adjust when AutoOps detects this event and presents the insight. Refer to AutoOps event settings for details.
The default customization settings are:
| Setting | Type | Default |
|---|---|---|
| Disk-to-memory ratio threshold for frozen-tier data nodes | Integer | 1000 |
The following is an example of what you might see when this insight is triggered. Real insights use live data and links from your deployment or cluster.
The memory-to-disk ratio is higher than needed on the following frozen nodes: es-data-01 and es-data-02.
For example, on es-data-01, allocated memory is 64 GB and allocated disk is 10240 GB, a memory-to-disk ratio of 1:160. A common target ratio for frozen nodes is 1:160.
AutoOps shows different recommendations depending on how their conditions match your deployment or cluster.
Reduce node memory capacity
Condition: Always shown for this insight.
To improve your memory-to-disk ratio, move frozen nodes from 64 GB to 16 GB by selecting instances with less memory.
Reduce data node count
Condition: Always shown for this insight.
Reduce the number of data nodes and increase disk allocated per remaining node accordingly.
The memory-to-disk ratio compares heap size to disk capacity on a node. When the ratio is higher than the target for that role or tier, you pay for RAM that the disk-backed workload does not need.
Frozen tiers rely heavily on searchable snapshots and local cache. Search is often I/O-bound, so surplus heap rarely improves latency enough to justify the cost.
If the condition persists, cloud or license cost stays elevated without a matching gain in search or indexing capacity.