Cluster Alarms¶
HPE Morpheus Software continuously monitors HVM cluster health and raises alarms when thresholds are exceeded or critical conditions are detected. Alarms appear in the cluster detail page and in the Operations > Health section.
Alarms are evaluated every 5 minutes during the cluster sync cycle. Threshold-based alarms use a sustained average over a configurable time window (default: 15 minutes) to avoid alerting on brief spikes.
Alarm Types¶
Host Alarms¶
Alarm |
Severity |
Description |
|---|---|---|
Host CPU Usage Warning |
Warning |
Host CPU usage is averaging above the warning threshold (default: 85%) over the monitoring window |
Host CPU Usage Critical |
Critical |
Host CPU usage is averaging above the critical threshold (default: 95%) over the monitoring window |
Host Memory Usage Warning |
Warning |
Host memory usage is averaging above the warning threshold (default: 90%) over the monitoring window |
Host Memory Usage Critical |
Critical |
Host memory usage is averaging above the critical threshold (default: 95%) over the monitoring window |
Host Memory Over-Committed |
Warning |
Total allocated VM memory exceeds the host’s physical memory capacity (adjusted by overcommit percent). See VM Placement & Scheduling for overcommit configuration. |
Host Unreachable |
Error |
Host has not reported in over 5 minutes and may be offline or experiencing network issues |
Multiple Heartbeat Datastores Detected |
Warning |
More than one datastore was configured as a heartbeat datastore. The system has auto-selected one. Review the Datastores tab to verify the selection. |
Datastore Alarms¶
Alarm |
Severity |
Description |
|---|---|---|
Datastore Usage Warning |
Warning |
Datastore capacity usage exceeds the warning threshold (default: 80%) |
Datastore Usage Critical |
Critical |
Datastore capacity usage exceeds the critical threshold (default: 90%) |
Datastore Offline |
Error |
Datastore is offline and not accessible from any host |
VM Alarms¶
Alarm |
Severity |
Description |
|---|---|---|
VM CPU Usage Warning |
Warning |
VM CPU usage is averaging above the warning threshold (default: 85%) over the monitoring window |
VM CPU Usage Critical |
Critical |
VM CPU usage is averaging above the critical threshold (default: 95%) over the monitoring window |
VM Memory Usage Warning |
Warning |
VM memory usage is averaging above the warning threshold (default: 90%) over the monitoring window |
VM Memory Usage Critical |
Critical |
VM memory usage is averaging above the critical threshold (default: 95%) over the monitoring window |
Operational Alarms¶
Alarm |
Severity |
Description |
|---|---|---|
Unable to Remove Host |
Warning |
A host removal operation could not safely complete. Follow Adding and Removing Hosts; do not manually edit cluster membership, and contact HPE Support if safe isolation or ring departure cannot be established. |
Default Thresholds¶
Metric |
Warning |
Critical |
Notes |
|---|---|---|---|
Host CPU |
85% |
95% |
Sustained average over window |
Host Memory |
90% |
95% |
Sustained average over window |
Host Overcommit |
100% |
— |
Total allocated vs physical capacity |
Datastore Usage |
80% |
90% |
Real-time capacity check |
VM CPU |
85% |
95% |
Sustained average over window |
VM Memory |
90% |
95% |
Sustained average over window |
Alarm Behavior¶
Monitoring Window¶
Threshold-based alarms (CPU, memory) use a rolling time window (default: 15 minutes) of averaged metrics from OpenSearch. This prevents short spikes from triggering false alarms.
Clear Margin¶
Alarms automatically clear when the metric drops below the threshold minus a clear margin (default: 10 percentage points). For example, a CPU warning alarm triggered at 85% will clear when usage drops below 75%. This hysteresis prevents alarm flapping.
Idempotent Evaluation¶
The alarm system is idempotent — if an alarm is already active, re-evaluating the same condition does not create duplicate alarms. Similarly, clearing an already-cleared alarm is a no-op.
Customizing Thresholds¶
Alarm thresholds can be overridden at the cluster level through the cluster configuration. Custom thresholds are merged with the defaults — you only need to specify the values you want to change.
Datastore-level threshold overrides are also supported, allowing different capacity warning levels for different datastores within the same cluster.
Note
Alarms can be disabled entirely for a cluster by setting enabled: false in the alarm thresholds configuration.
Viewing and Acknowledging Alarms¶
Alarms are visible in:
Cluster detail page — Summary panel shows active alarm count and status
Operations > Health — Global view of all active alarms across the appliance
To acknowledge an alarm:
Navigate to the alarm in Operations > Health
Click the alarm to view details
Click Acknowledge to mark it as seen
Acknowledged alarms remain visible but no longer contribute to the cluster’s overall health status indicator. Alarms automatically clear when the underlying condition resolves.