I'm opening a case on this now. At one site have 2 of these generated and there are 0 items in the “All Monitors” section that pertain to the items in the alert.
Even tried upgrading to 6.82 version cleared alerts and they still come back.
Self-Monitoring Failures for '***-vcenter:CLUSTER_COMPUTE_RESOURCE.MemoryUsage': Data Collection (2 of 4 failed). See vmware.log for more details
Self-Monitoring Failures for '***-vcenter:CLUSTER_COMPUTE_RESOURCE.CPUusage': Data Collection (2 of 4 failed). See vmware.log for more details
and in log:
Mar 16 09:11:36:379 [Data Collector - ***-vcenter, vmware] SENDING_ALARM: Self-Monitoring Failures for '***-vcenter:CLUSTER_COMPUTE_RESOURCE.CPUusage': Data Collection (2 of 4 failed). See vmware.log for more details
Mar 16 09:11:36:379 [Data Collector - ***-vcenter, vmware] SENDING_ALARM: Self-Monitoring Failures for '***-vcenter:CLUSTER_COMPUTE_RESOURCE.MemoryUsage': Data Collection (2 of 4 failed). See vmware.log for more details
Mar 16 09:11:36:379 [Data Collector - ***-vcenter, vmware] ===== Self-Monitoring Alarm Failures: 2 alarm(s) sent for resource: ***-vcenter =====
Note 1: These Self-Monitoring Alarm Failures are aggregated for an element.metric type per resource. Individual failure details or related exceptions should proceed this log entry.
Note 2: 'Monitor Correlation' failures occur when a monitor does not find it's specific element in the inventory, or no metric value is available for the element.
With static monitors and changing inventory, these are sometimes expected and may be transitory.
Note 3: The failure count of 'Data Collection' failures often correlate with Monitor Correlation failures.
When there are only 'Data Collection' failures, or when they exceed 'Monitor Correlation' failures, that usually indicates a problem in collecting that metric value.
Some metric values are only available with additional system administration.
Some metric values are only available for specific element types. For instance one type of storage might have a metric, while another does not.
Generally it is desirable to understand 'Data Collection' failures for desired metrics, and sometimes the probe needs to be tuned for them