Azure PostgreSQL Flexible Server: SkuNotAvailable in UAE North despite available quota

Ed 0 Reputation points
2026-11-13T22:14:35.7833333+00:00

Azure PostgreSQL Flexible Server — Repeated SkuNotAvailable in UAE North

We are experiencing repeated provisioning failures when attempting a point-in-time restore (PITR) of an existing Azure Database for PostgreSQL Flexible Server in the UAE North region.

Environment

Region: UAE North

PostgreSQL version: 16

Source server: Standard_B1ms / Burstable, Availability Zone 1

Source networking: Private VNet integration, public network access disabled

Restore target: Temporary PostgreSQL Flexible Server in the same region and private VNet

Subscription: Active Pay-As-You-Go

PostgreSQL vCore usage: 2 of 384

Burstable-family vCore usage: 2 of 40

Problem

Azure repeatedly returns:

HTTP 409 Conflict — SkuNotAvailable

"Specified SKU is not supported in this region. Please choose a different SKU."

We have attempted the following configurations:

Configuration Outcome
Standard_B1ms / Burstable Failed
Standard_B2s / Zone 1 Succeeded once, then repeatedly failed
Standard_B2s / Azure-selected zone Failed
Standard_B2ms / Zone 2 Failed

The regional SKU capability API advertises these configurations as supported. ARM template validation also succeeds.

Importantly, Standard_B2s / Zone 1 successfully provisioned a restored server on November 13, 2026. That deployment subsequently failed due to an unrelated Container Apps Job naming issue, which has since been corrected.

Subsequent PostgreSQL provisioning attempts failed even with the same SKU, tier, zone, source and network configuration.

Troubleshooting already completed

Confirmed PostgreSQL resource provider is registered.

Confirmed subscription is active.

Confirmed sufficient vCore quota.

Checked Azure Policy; no applicable deny policy identified.

Verified delegated subnet and private DNS configuration.

Verified restore request matches documented PostgreSQL Flexible Server PITR ARM requirements.

Tested alternative SKUs and availability-zone selections.

Verified temporary server deletion and name availability.

Checked Azure Service Health; no relevant active incident was identified.

Confirmed deployments are accepted by ARM but fail asynchronously during PostgreSQL allocation.

Multiple provisioning attempts failed with HTTP 409 SkuNotAvailable. Detailed Azure deployment correlation IDs and diagnostic evidence are available upon request through a private Microsoft support channel.

Questions for Microsoft engineers

Could this indicate a PostgreSQL Flexible Server capacity allocation issue or a subscription-specific resource-provider restriction in UAE North, despite available quota?

Why would the same B2s/Zone 1 PITR configuration succeed once and then repeatedly fail?

Could deleted-server metadata or backend tombstone state affect subsequent provisioning attempts using the same server name?

Is there a recommended SKU and availability-zone combination that can currently be allocated?

Would switching to General Purpose (for example, Standard_D2ds_v5) help, and is direct PITR from a Burstable B1ms source to that tier supported?

Is there an alternative supported approach to perform this managed PITR without repeated failed allocation attempts?

We only require a short-lived Development restore server for database migration and security verification. It will be deleted after testing.

Any technical guidance or assistance from Microsoft PostgreSQL engineers would be greatly appreciated.

Azure Database for PostgreSQL
0 comments No comments

1 answer

Sort by: Oldest
  1. AI answer

    2026-11-13T22:15:51.7033333+00:00

    The next supported action is to open an Azure PostgreSQL Flexible Server Region access request for UAE North, because the documented remedies—different SKU and availability zone—have already failed and the allocation error occurs after ARM accepts the deployment.^1^

    What the evidence indicates

    SkuNotAvailable is distinct from a quota-exceeded error. Microsoft documents that the message can mean either:

    • The SKU is unsupported in that region, or
    • The SKU is temporarily unavailable there.

    Capacity can also be availability-zone-specific.^1^

    That fits the observed pattern: B2s/Zone 1 allocated once, while later requests with the same configuration failed asynchronously. SKU capability output and successful ARM validation describe valid configurations; they do not establish that capacity is allocatable for this subscription at a particular time.

    The available public information does not establish whether this specific UAE North failure is regional capacity or a subscription-scoped backend restriction. Microsoft support must inspect the failed allocation records to distinguish them. No public UAE North incident or currently allocatable SKU/zone combination is identified in the available documentation.

    Open the correct support request

    In Azure portal → Help + support → Create a support request:

    1. Select Azure Database for PostgreSQL Flexible Server.
    2. Choose the capacity/region-access path—documented as a Region access request.
    3. Specify:
      • Region: UAE North
      • Requested tier/SKU and vCores
      • PostgreSQL version 16
      • PITR operation
      • Each attempted SKU and zone
      • UTC timestamps
      • Deployment and correlation IDs
      • The successful B2s/Zone 1 allocation time
      • Confirmation that quota, RP registration, policy, subnet, DNS, and Service Health were already checked
    4. Ask support to confirm:
      • Current capacity by SKU and zone
      • Whether the subscription has a regional or SKU allocation restriction
      • Whether any failed/deleted target resource remains in backend state
      • A currently allocatable Burstable SKU/zone for PITR

    Microsoft’s capacity guide explicitly directs unresolved provisioning failures to a Region access request and asks that the required vCores be included.^1^

    General Purpose cannot be selected during this PITR

    A direct PITR from Standard_B1ms to Standard_D2ds_v5 is not supported as part of the restore operation. PITR creates a new server with the source server’s:

    • Pricing/compute tier
    • Compute generation and vCore count
    • Storage size
    • Backup retention and redundancy settings

    The restore wizard specifically marks the compute tier and compute size as unchangeable during restoration. Compute can be scaled after the restored server has been created.^2^^3^

    Therefore, General Purpose helps only through one of these paths:

    1. Successfully restore as B1ms, then scale the new server to General Purpose.
    2. Create a separate General Purpose server and perform a logical migration using pg_dump/pg_restore.

    Submitting a nominal B2s/B2ms restore configuration does not change this documented PITR restriction; support should verify why those requests were accepted or allocated in this case.

    Deleted target names and backend state

    Ordinary PITR always creates a new server and does not overwrite the source.^4^

    There is no documented basis for concluding that a deleted temporary target normally causes later SkuNotAvailable responses. A backend inconsistency is possible in exceptional cases: one PostgreSQL engineering case required manual cleanup after ARM and the backend became unsynchronized.^5^ That case involved a resource that remained active and billable after deletion, so it does not prove a tombstone issue here.

    Use a fresh target name for any further controlled test, but avoid repeated retries until support confirms capacity. Name availability alone cannot exclude backend state; support can check the prior target resource IDs.

    Alternatives if the restore is urgent

    Managed PITR alternatives are limited:

    • Same-region PITR must use the source compute configuration.
    • Geo-restore is available only when the source was created with geo-redundant backups; it creates the target in the paired region.^4^
    • For a private-access source, same-region PITR can use the same or another VNet, but it cannot switch to public access.

    If a historical timestamp is not essential, create an allocatable General Purpose server and logically copy the current data. Use PostgreSQL 16 or newer client utilities and include roles separately because pg_dump does not export global roles.^6^

    pg_dumpall -r --no-role-passwords \
      -h source.postgres.database.azure.com -U admin > roles.sql
    
    pg_dump -Fd -j 2 appdb \
      -h source.postgres.database.azure.com -U admin -f appdb.dump
    
    psql -f roles.sql -h target.postgres.database.azure.com -U admin
    pg_restore -Fd -j 2 -d appdb appdb.dump \
      -h target.postgres.database.azure.com -U admin 2> errors.log
    

    This produces a logical copy at dump time, not a managed PITR to an earlier timestamp.


    References

    1. Resolve capacity errors - Azure Database for PostgreSQL
    2. Restore to latest restore point
    3. Limits in Azure Database for PostgreSQL flexible server
    4. Backup and restore in Azure Database for PostgreSQL flexible server
    5. Unable to Access PostgreSQL Flexible Servers – Request for Restoration to Supported SKU/Region - Microsoft Q&A
    6. Migrate your PostgreSQL database by using dump and restore
    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.