ServiceUnavailable creating Linux container group with K80 GPU

Question

I am attempting to run a Docker image in a new container group, using a K80 GPU.

The YAML file defining the container group includes the following (plus other stuff):

location: westus2
properties:
        requests:
          cpu: 1.0
          memoryInGB: 16.0 
          gpu:
            count: 1
            sku: K80

I use the az container create command and get the following error:

Code: ServiceUnavailable
Message: The requested resource is not available in the location 'westus2' at this moment. Please retry with a different resource request or in another location. Resource requested: '1' CPU '8' GB memory 'Linux' OS '1' 'K80' GPU

I've tried several varying the memory and CPU requirements, and switching the location to eastus, with the same results.

If I omit the GPU requirement, it succeeds, except that the resulting container quits once it realizes it has no GPU.

My quota for K80 GPUs is over 100, and none are in use.

What else should I try?

Answer 1

@Peter Venable

There appears to be some resource constraints for the K80 container group in westus2. More information about this error can be found here.

I checked and can see there is currently available capacity for K80 in both North Europe and West Europe. If possible, you can deploy to these regions. Otherwise, you will need to change to another size or wait for the deployment.

---

Please don’t forget to "Accept the answer" and “up-vote” wherever the information provided helps you, this can be beneficial to other community members.