PostgreSQL Flexible Server stuck in "Updating" since 2026-11-19 after scaling from General Purpose to Burstable B1ms (East US 2)

Quiet Harbor 0 Reputation points
2026-11-20T16:19:56.3033333+00:00

Our Azure Database for PostgreSQL flexible server has been in the Updating state since 2026-11-19 at 14:33 UTC, after a compute scale-down. The operation neither finishes nor fails. Our subscription has the Free support plan, so we can't open a technical support request.

Server

  • Region and zone: East US 2, zone 2
  • PostgreSQL 18, 32 GB storage (auto-grow off), high availability off
  • Private access (VNet integration with a delegated subnet), public access disabled
  • Current compute: General Purpose Standard_D2ds_v5
  • Requested compute: Burstable Standard_B1ms

What happened

  • 2026-11-19 at 14:33:51 UTC we sent the scale change (Terraform azurerm provider, in-place update of sku_name). The activity log shows Started and then Accepted (HTTP 202) for Microsoft.DBforPostgreSQL/flexibleServers/write, correlation ID cbdf7aa8-fb50-b7d6-7b80-801297516c52.
  • Terraform's 60-minute timeout ran out, but the operation kept running in Azure. The activity log has no Succeeded or Failed event for it.
  • Since then the activity log has shown a flexibleServers/read event with status Running and the same correlation ID every minute.
  • 2026-11-19 at 20:50 UTC we retried the same change. Azure refused it with ServerIsBusy ("Cannot complete operation while server '<server>' is busy processing another operation"), so the first operation is still running.
  • Latest check, 2026-11-20 at 13:45 UTC: az postgres flexible-server show still returns state Updating and SKU Standard_D2ds_v5.
  • Our dev app keeps running on the server, with no database errors in its logs.

Background

  • On 2026-11-15, creating this server as Burstable B1ms in East US 2 failed with SkuNotAvailable ("Specified SKU is not supported in this region"), both in the zone Azure picked and in zone 2. We created it as General Purpose as a workaround.
  • az postgres flexible-server list-skus -l eastus2 lists Standard_B1ms in zones 2, 3 and 4.
  • Quota isn't the blocker: the usages API shows standardBSFamily with limit 40 and 0 used in eastus2. For that reason the portal's quota request fails.

Questions

  1. Can someone from the PostgreSQL team check this operation and either complete it or roll it back to General Purpose?
  2. Is Burstable B1ms capacity available in East US 2 zone 2 now? If not, which zone has it? The server has to stay in this region because of the VNet.
  3. Is there a supported way for us to cancel a stuck scale operation?

I can share the subscription ID and the server's resource ID in a private message.

Azure Database for PostgreSQL
0 comments No comments

1 answer

Sort by: Oldest
  1. AI answer

    2026-11-20T16:23:56.5866667+00:00

    Stop submitting updates to this server and request Microsoft backend recovery, because the persistent Updating/ServerIsBusy state cannot be cleared through a documented customer-facing PostgreSQL Flexible Server cancellation operation.

    Recover the stuck server

    Similar Flexible Server scale operations that remained in Updating were recovered only after Microsoft restarted unhealthy backend components or containers.^1^^2^ That is a close symptom match, but only Microsoft can determine whether the same backend fault affects this server.

    1. Do not retry the scale, restart, stop, or another write operation.
      The second request already returned ServerIsBusy; additional writes will not supersede the accepted operation.
    2. Check the server’s Resource Health page:
      • Azure portal → PostgreSQL flexible server → Resource health.
      • If it shows Unavailable due to a platform event, select the support option from that page. Microsoft states that an unavailable resource affected by a platform event can open a support request without a support agreement.^3^
      • If it shows Available, select Report incorrect health status and provide the start time, operation, and correlation ID. This submits the health signal for reassessment; it does not guarantee that a technical support case will be created.
    3. Check Azure portal → Service Health → Service issues and Health advisories.
      Service Health is subscription-specific and can expose incidents too narrow to appear on the public Azure status page.^4^ If an applicable incident exists, follow it there; a free recovery ticket can be raised for up to 72 hours after a service disruption is mitigated.^5^
    4. If Resource Health remains Available and no incident applies, the official remaining paths are:
      • Post a sanitized Microsoft Q&A case and request moderator/product-team escalation. Include the UTC timeline, region/zone, old and requested SKUs, Updating state, ServerIsBusy, and correlation ID. Do not post the subscription ID or resource ID publicly.
      • Ask the subscription owner or tenant administrator whether the organization has an Azure support agreement that can be linked to the subscription.
      • If neither route is available, obtain an eligible paid support plan for guaranteed technical case handling. The support request should ask Microsoft to complete or roll back the existing operation and return provisioning state to Ready, not initiate another scale request.

    Burstable availability in East US 2

    The current published Azure Database for PostgreSQL Flexible Server regional table marks Burstable as unsupported in East US 2.^6^ Therefore, no East US 2 availability zone can presently be recommended for Standard_B1ms from public documentation.

    az postgres flexible-server list-skus showing B1ms in zones 2–4 does not prove that PostgreSQL can allocate that SKU at the moment. Microsoft documents that SKU availability can differ by region and can also be temporarily unavailable; for SkuNotAvailable, the supported response is to select another SKU type.^7^

    Only Microsoft can confirm live, subscription-specific, zone-level capacity. Given both the November 15 SkuNotAvailable result and the regional support table, keep the server on a supported General Purpose SKU after recovery rather than retrying B1ms. The existing VNet requirement constrains the region, but it does not establish that Burstable capacity exists in any East US 2 zone.

    Cancellation and Terraform state

    No supported customer-facing cancellation mechanism is documented for a PostgreSQL Flexible Server compute-scaling operation. The documented command starts a scale operation but provides no corresponding compute-scale cancel command.^8^ Migration cancellation commands apply only to PostgreSQL migration workflows and must not be used for compute scaling.

    Terraform’s timeout stopped waiting locally; it did not cancel the Azure operation. Leave the configured SKU unchanged until Azure returns the server to Ready. After Microsoft recovery:

    1. Run az postgres flexible-server show and confirm state is Ready.
    2. Confirm which SKU Azure retained.
    3. Refresh Terraform state.
    4. Set sku_name to that recovered SKU before the next plan, preventing Terraform from immediately retrying B1ms.

    Compute scaling normally provisions replacement compute, attaches the existing storage, and restarts the database before accepting connections.^9^ The application continuing to run suggests the original General Purpose data plane remains active, but it does not mean the control-plane operation has completed.


    References

    1. Azure PostgreSQL server is stuck in 'updating' state - Microsoft Q&A
    2. PostgreSQL Flexible Server stuck on change Compute and Storage tier operation - Microsoft Q&A
    3. Azure Resource Health FAQ
    4. What is Azure Service Health?
    5. What to do during an Azure service disruption
    6. What is Azure Database for PostgreSQL flexible server?
    7. Resolve capacity errors for Azure Database for PostgreSQL flexible server
    8. Scale compute in Azure Database for PostgreSQL flexible server
    9. Overview of business continuity in Azure Database for PostgreSQL flexible server
    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.