Request quota limit increases for some of the quotas for GPU enabled SaaS Linux runners
For each GCP project we create in the create projects issue we also need to request quota increases for the "usual things".
Basing on the recent work done for saas-linux-large-amd64
and saas-linux-medium-amd64
runners and assuming that we will handle the same maximum capacity per the GCP project as there, we will definitely need to request upgrades for:
- Read requests per minute - 4500 (from the default 1500)
- Heavy-weight read requests per minute - 2250 (from the default 750)
We will also need to increase the CPUs
quota. The default limit is 2400 and with bigger machine types we will use definitely more. Assuming that the maximum number of VMs we will handle in each GCP project will be 1500, we will need to request 6000 for the saas-linux-large-amd64-gpu
shard (using the n1-standard-4
machine type which has 4 vCPUs).
Finally we will most probably need to request increase a limit for the GPU quota. Which quota and how big limit we will need will depend on the results of discussion happening in https://gitlab.com/gitlab-com/gl-infra/reliability/-/issues/16345.
Work checklist
- Request quota limit increase:
-
gitlab-r-saas-l-l-amd64-gpu-1
Thank you for submitting Case # (ID:44069859) to Google Cloud Platform support for the following quotas: Change NVIDIA T4 GPUs - us-east1 from 16 to 1,500 Change CPUs - us-east1 from 2,400 to 6,000 Change Heavy-weight read requests per minute per region - us-east1 from 750 to 2,250 Change Read requests per minute from 1,500 to 4,500 Change Heavy-weight read requests per minute from 750 to 2,250
-
gitlab-r-saas-l-l-amd64-gpu-2
Thank you for submitting Case # (ID:57e63725347f43c096) to Google Cloud Platform support for the following quotas: Change NVIDIA T4 GPUs - us-east1 from 16 to 1,500 Change CPUs - us-east1 from 2,400 to 6,000 Change Heavy-weight read requests per minute per region - us-east1 from 750 to 2,250 Change Read requests per minute from 1,500 to 4,500 Change Heavy-weight read requests per minute from 750 to 2,250
-
gitlab-r-saas-l-l-amd64-gpu-3
Thank you for submitting Case # (ID:5460116c56a54ba181) to Google Cloud Platform support for the following quotas: Change CPUs - us-east1 from 2,400 to 6,000 Change Heavy-weight read requests per minute from 750 to 2,250 Change NVIDIA T4 GPUs - us-east1 from 16 to 1,500 Change Read requests per minute from 1,500 to 4,500
-
gitlab-r-saas-l-l-amd64-gpu-4
Thank you for submitting Case # (ID:1b16fb024c224d3e84) to Google Cloud Platform support for the following quotas: Change CPUs - us-east1 from 2,400 to 6,000 Change NVIDIA T4 GPUs - us-east1 from 16 to 1,500 Change Read requests per minute from 1,500 to 4,500 Change Heavy-weight read requests per minute from 750 to 2,250
-
gitlab-r-saas-l-l-amd64-gpu-5
Thank you for submitting Case # (ID:5b3c2d19c8f24d0b81) to Google Cloud Platform support for the following quotas: Change CPUs - us-east1 from 2,400 to 6,000 Change NVIDIA T4 GPUs - us-east1 from 16 to 1,500 Change Read requests per minute from 1,500 to 4,500 Change Heavy-weight read requests per minute from 750 to 2,250
-
- Confirm that the limit was increased:
-
gitlab-r-saas-l-l-amd64-gpu-1
-
Change CPUs - us-central1 from 2,400 to 6,000 -
Change NVIDIA T4 GPUs - us-central1 from 16 to 1,500 -
Change Read requests per minute from 1,500 to 4,500 -
Change Heavy-weight read requests per minute from 750 to 2,250
-
-
gitlab-r-saas-l-l-amd64-gpu-2
-
Change CPUs - us-central1 from 2,400 to 6,000 -
Change NVIDIA T4 GPUs - us-central1 from 16 to 1,500 -
Change Read requests per minute from 1,500 to 4,500 -
Change Heavy-weight read requests per minute from 750 to 2,250
-
-
gitlab-r-saas-l-l-amd64-gpu-3
-
Change CPUs - us-central1 from 2,400 to 6,000 -
Change NVIDIA T4 GPUs - us-central1 from 16 to 1,500 -
Change Read requests per minute from 1,500 to 4,500 -
Change Heavy-weight read requests per minute from 750 to 2,250
-
-
gitlab-r-saas-l-l-amd64-gpu-4
-
Change CPUs - us-central1 from 2,400 to 6,000 -
Change NVIDIA T4 GPUs - us-central1 from 16 to 1,500 -
Change Read requests per minute from 1,500 to 4,500 -
Change Heavy-weight read requests per minute from 750 to 2,250
-
-
gitlab-r-saas-l-l-amd64-gpu-5
-
Change CPUs - us-central1 from 2,400 to 6,000 -
Change NVIDIA T4 GPUs - us-central1 from 16 to 1,500 -
Change Read requests per minute from 1,500 to 4,500 -
Change Heavy-weight read requests per minute from 750 to 2,250
-
-