Azure compute creation failure on Personal Compute policy in Sweden Central/West Europe

Question

Azure compute creation failure on Personal Compute policy in Sweden Central/West Europe

Hussain Doctor 0

Hello,

I am unable to create or start compute in my Azure Databricks workspace. This has been failing for several days. The affected compute is in Azure Sweden Central, using the Personal Compute policy/Unrestricted. It is a classic single-node cluster with DBR 19.x-scala2.13, access mode Standard, and current node type Standard_D4ds_v4. The failure message shows quota/capacity-related errors, including: QuotaExceeded: Operation could not be completed... and The VM size you are specifying is not available, and the cluster sits in PENDING / TERMINATED.

I already checked Azure quota and there appears to be enough quota in the subscription/region, so this looks inconsistent with the actual availability/error being returned. I also tried multiple allowed node types under the same policy, including Standard_D4ds_v5 and Standard_D4ds_v4, and previously Standard_DS3_v2 was shown as unsupported in the subscription. Please investigate whether this is an Azure capacity issue, stale quota validation, or a Databricks-side provisioning problem for this workspace/policy. This is urgent because my access/trial expiry is near and I am blocked from deploying any compute.

Peram Venkateswara Reddy 255 Reputation points Microsoft External Staff Moderator

2026-08-04T11:52:46.3166667+00:00
Hi @Hussain Doctor , Hope you are doing well. Hi,

Thanks for the detail - that's enough to identify that this is Azure capacity (stockout), not your quota. Databricks reports this as CLOUD_PROVIDER_RESOURCE_STOCKOUT, and the documented example is nearly identical to yours: "Following SKUs have failed for Capacity Restrictions… is currently not available in location 'westeurope'." Regional vCPU quota can look completely healthy while a specific SKU has no physical capacity — quota increases won't help.

Do this first — enable flexible node types**.** This is the documented fix for this exact error. Ask your workspace admin to enable it (Admin settings → Compute), then create a new compute resource — existing clusters aren't retrofitted. Databricks will then automatically fall back to compatible instance types instead of failing.

If you can't enable it, in this order:

Retry the launch — capacity fluctuates; off-peak hours help.

Try a different VM family (not just v4→v5 within Dds — e.g. Standard_E or Standard_F sizes).

Set Availability to on-demand rather than zone-pinned so Databricks can place across zones.

If Sweden Central stays constrained, try West Europe or another nearby region.

Check the Azure Service Health dashboard for regional capacity incidents.

One thing worth double-checking, given your trial: quota is enforced per VM family, not just Total Regional vCores. In Portal → Subscription → Usage + quotas, filter to Standard DDSv4/DDSv5 Family vCPUs specifically. Trial and student subscriptions frequently sit at 0–4 for individual families even when the regional total looks fine — and DS3_v2 already showing as unsupported in your subscription points this way. If that's the case, upgrading to Pay-As-You-Go removes the constraint immediately, which matters given your expiry date.

Also consider serverless compute - it bypasses classic VM provisioning entirely and would unblock you today if it's available in your workspace.

References

Classic compute termination error codes — CLOUD_PROVIDER_RESOURCE_STOCKOUT

Improve compute launch reliability using flexible node types

If it still fails after enabling flexible node types, share the full azure_error_message from the event log and I'll narrow it further.

Kind Regards,
Microsoft Support Team.
Peram Venkateswara Reddy 255 Reputation points Microsoft External Staff Moderator

2026-08-05T08:18:07.3+00:00

Hi @Hussain Doctor , Hope you are doing well. Following up on my previous comment regarding your query. Please let us know if it resolved your query.

Kind Regards,
Microsoft Support Team.

1 answer

Your answer

Peram Venkateswara Reddy 255 Reputation points Microsoft External Staff Moderator

2026-08-05T08:18:07.3+00:00

Hi @Hussain Doctor , Hope you are doing well. Following up on my previous comment regarding your query. Please let us know if it resolved your query.

Kind Regards,
Microsoft Support Team.

Answer 1

For this Azure Databricks scenario, the supported causes are quota limits, Azure permission or policy restrictions, and regional VM capacity/stockout.

Based on the reported errors:

Check the exact cluster event log error code first.
- In the compute event log, look for whether the failure maps to AZURE_QUOTA_EXCEEDED_EXCEPTION or AZURE_OPERATION_NOT_ALLOWED_EXCEPTION.
- AZURE_QUOTA_EXCEEDED_EXCEPTION means the requested VM family quota would be exceeded.
- AZURE_OPERATION_NOT_ALLOWED_EXCEPTION means the Azure operation failed because of permission restrictions, policy violations, or account limitations.
Re-verify quota at the VM family level, not only overall regional quota.
- In Azure portal, go to Subscriptions > Usage + quotas.
- Confirm the quota for the exact VM family used by the node type, and review current usage from other VMs or clusters in the same subscription.
- The documented fix for quota failures is to request a quota increase, terminate unused clusters, or use a different VM type with available quota.
Treat the “VM size is not available” message as a regional capacity issue if quota looks sufficient.
- The documented Q&A guidance for this pattern is that the SKU can be temporarily unavailable in a region due to stockout/capacity constraints, even when quota exists.
- Since multiple node types in Sweden Central and West Europe are failing, test another allowed VM size or another region if available to isolate whether the issue is regional capacity.
Check Azure-side restrictions that can look like quota problems.
- Review Azure activity logs for the failed deployment.
- Check service principal permissions on the resource group.
- Verify Azure policies are not blocking the operation.
- Check the subscription state and status.
- Review recent permission or policy changes.
Confirm the policy path is not the blocker.
- Personal Compute is available by default to all users for single-machine compute resources.
- If a user does not have unrestricted cluster creation, compute creation is limited by granted policies.
- If the policy dropdown is visible and Personal Compute is selectable, that indicates policy access exists; the documented failure patterns for the reported messages still point to quota, Azure restrictions, or capacity.
If custom container settings are involved, verify they are not contributing.
- If the compute uses a custom Docker image, image pull or container creation failures can also terminate compute.
- If no custom container is configured, this can be ruled out.

What is supported by the available guidance for this case:

A Databricks-side provisioning defect for this specific workspace or Personal Compute policy is not established by the provided documentation.
The documented causes that match the reported symptoms are:
- VM family quota exhaustion
- Azure permission/policy/account restrictions
- Regional VM SKU capacity unavailability

Most direct next path:

Open the compute event log and capture the exact termination code.
In Azure portal, check Subscriptions > Usage + quotas for the exact VM family.
Check Azure activity logs for the failed create/start attempt.
If quota is available and activity logs show SKU unavailability, retry with another allowed VM size or region.
If activity logs show permission/policy/account restriction, correct that Azure-side issue.
If the failure remains after those checks, contact Azure support for quota/permission configuration assistance.