I'm trying to provision a V100 GPU on GCP using SkyPilot and the service is hanging at the PROVISIONING stage. I have the necessary quota for a V100 and I'm using a service account with 'Editor' and 'Service Config Editor' permissions, in addition to the Medium Permissions recommended by SkyPilot. Is there something I'm missing that's causing this issue?
Gergő Szabó
Asked on Apr 07, 2024
It seems like you've checked the necessary permissions and quotas, but you're still facing provisioning issues. A common cause for such issues could be regional availability or temporary resource shortages. Make sure to check the provision logs for detailed error messages, which you can access using !sky serve logs serving-test --controller
. In your case, the logs indicated a quota issue related to SSD limits, which was the root cause of the provisioning hang-up.