Alarms in the DX SaaS platform remain in an "OPEN" status even after the underlying metric value has recovered and no longer meets the major or critical threshold.
metric_state transitioned from "ACTIVE" to "UNKNOWN" but did not transition back to "ACTIVE" or "CLOSED" when data returned.The issue is related to defect # DE667937
The root cause is a synchronization glitch within the metricalert service pods. When multiple metricalert pods are restarted (typically to remediate performance lag), a race condition or state mismatch can occur. If an alarm cycle is being processed exactly when the pods are restarting, the alarm can get stuck in an "UNKNOWN" state and fail to trigger the subsequent "CLEAR" event once the metric recovers.
This issue is fixed in 26.8.2 and onward versions.
Workaround: Manual restart of metric alert pod instances