What to do when Quota page shows CPUs per VM family is at 14 but I have no running instances

[rholowczak]

Folks:

I have been doing some testing setting up Dataproc clusters. I have run into so many different quota issues it is ridiculous. CPUS_PER_VM_FAMILY, IN_USE_ADDRESSES, Hyperdisk Balanced Capacity, VM instances and CPUs (all regions) all had to be increased.

Now I have deleted all clusters. I see no running VMs in any region.

However, my utilization of CPUs per VM family is currently "stuck" at 14 and will not go down.  So now I am not able to launch any more clusters even though I don't have any clusters at all at this point.

I have two small VM Instances (based on E2 machine types) that are both stopped.

Is there any way to trigger the Quota system to "Refresh" its internal records?

Yes I have refreshed the Console web page many times over the last 2 hours.

Also when I run: gcloud compute regions describe us-central1

and I run gcloud compute project-info describe --project 

I cannot see the CPUS_PER_VM_FAMILY quota at all.

It would be terrific if all of this could be made consistent (especially the naming).

Any suggestions on getting GCP to "refresh" its internal records?

Thanks

Rich H.

[nmagcalengjr]

Hi @rholowczak,

Welcome to Google Cloud Community!

Here are some basic troubleshooting steps you can follow:

1. Restart the Stopped VMs:

Even if the instances are stopped, GCP may not have fully released the allocated CPU resources, which can affect quota tracking. Refreshing the quota may help resolve this. You can try below commands,

gcloud compute instances start INSTANCE_NAME --zone=ZONE

gcloud compute instances stop INSTANCE_NAME --zone=ZONE

2. Modify and Revert Quota Requests:

Requesting a temporary quota decrease followed by an increase can prompt GCP to recalculate resource allocations, potentially resolving quota-related issues.

• Go to IAM & Admin → Quotas
• Locate CPUs per VM family (or any CPU-related quota)
• Click Edit Quota → Request a small decrease (e.g., from 14 to 10)
• Wait a few minutes
• Request the original quota back

3. Trigger a No-op Quota Request: 

If the quota field is missing in gcloud compute project-info describe, try running a dummy quota request or explicitly fetching quota details. Submitting a small quota change request, even if not needed, can sometimes trigger a backend refresh.

gcloud compute project-info describe --verbosity=debug

gcloud compute regions describe us-central1 --format=json

4. Try a Metadata Update:

Modifying project metadata, such as adding a temporary key, can sometimes trigger a quota recalculation. After a few minutes, you can remove the key to see if the quota updates.

gcloud compute project-info add-metadata --metadata temp-key=temp-value

gcloud compute project-info remove-metadata --keys=temp-key

If the quota remains stuck despite troubleshooting, contacting Google Cloud Support may be necessary. They can manually reset the quota, triggering a backend update for further investigation in your project.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

[rholowczak]

Hi. Thanks for all of these suggestions. I tried each one but there is no change.  Resource usage remains "sticky".  I ran a cluster yesterday with four  C4 CPUs. It ran for about 5 minutes and then I deleted the cluster.  Almost 24 hours later it still shows I am using four CPUs. I tried adjusting the quota up and down, I tried the temporary key thing, etc. Nothing changed the fact that Google thinks I am still using four of these CPUs.

Unfortunately, I can;t submit a support ticket as I am working from an educational credit account.

Is there any possible way to get word to someone within Google who works on the quota system to try and sort this out?