Your data. Anywhere you go.

New Relic for iOS or Android

Download on the App Store    Android App on Google play

New Relic Insights App for iOS

Download on the App Store

Learn more

Close icon

How to view remaining request capacity


I would like to build a dashboard/alert which shows me the sum of K8sContainerSample.cpuRequestedCores versus how many total cores are schedule-able in the cluster. I have not been able to figure this out.

For requested cores I can only get the latest values like this:
FROM K8sContainerSample SELECT latest(cpuRequestedCores) FACET containerID
But I am not able to sum the values.

For the total cores in the cluster, I am unsure how where to get this metric from.

Any pointers would be appreciated.


Hi @mike.luedke ,

One trick in order to sum value is to do the average * the number of instances.
In this case you could do :
FROM K8sContainerSample SELECT average(cpuRequestedCores)*uniqueCount(containerID)

The number of cores get reported from the underlying OS in the SystemSample event type.
To get the total request vs the total cores ; you would use something like:
from SystemSample,K8sContainerSample SELECT (average(cpuRequestedCores)*uniqueCount(containerID))/( average(numeric(coreCount))*uniqueCount(hostname) )



Thanks for this suggestion. Unfortunately the result seems to be way off from the actual values that we see in kube-state-metrics. I think the problem with this approach is that we lose too much fidelity when we are talking about thousands of containers.

Edit: I also realized the suggestion above averages all historical values, not just current values. That makes the calculation invalid.


You could restrict the timeframe so that averages are more accurate:
from SystemSample,K8sContainerSample SELECT (average(cpuRequestedCores)*uniqueCount(containerID))/( average(numeric(coreCount))*uniqueCount(hostname) ) since 2 minutes ago until 1 minute ago

Another approach is to represent the data as a timeserie so you can see the evolution:
from SystemSample,K8sContainerSample SELECT (average(cpuRequestedCores)*uniqueCount(containerID))/( average(numeric(coreCount))*uniqueCount(hostname) ) since 1 day ago TIMESERIES