0

We're using Bluemedora BindPlane to send on-prem utilization metrics to GCP.

In so doing, we're trying to create a GCP monitoring alert that will alert us if the CPU % for a server goes and stays above 95% for 15 minutes.

But we're getting an alert incident for each core.

Is anyone familiar enough with MQL such that they could offer how I could restructure the below such that I obtain an average across cores, rather than an incident per core?

fetch generic_node
| metric 'external.googleapis.com/bluemedora/linux_host/cpu/utilization'
| filter (resource.node_id == 'hdn-tst-app06')
| group_by [metric.core], [value_utilization_mean: mean(value.utilization)]
| every 15m
| condition value_utilization_mean > 95 '%'
| window 15m
theosophe74
  • 121
  • 1
  • 1
  • 5

1 Answers1

0

I would try to add a secondary aggregation, using another group_by.

| group_by [metric.core],
    [value_utilization_mean_mean: mean(value_utilization_mean)]
| group_by [],
    [value_utilization_mean_mean_mean: mean(value_utilization_mean_mean)]
| condition value_utilization_mean > 95 '%'
J.Vander
  • 1