Azure Database for PostgreSQL Flexible Server stuck in 'Starting' status

Question

Azure Database for PostgreSQL Flexible Server stuck in 'Starting' status

Alessandro Vernile 40

Following a manual stop operation, the server has been stuck in a 'Starting' status. I am unable to perform any management operations, and the Azure Portal Resource Health displays an "Unknown (Unplanned)" status.

Troubleshooting already performed:

I have attempted to restart/stop/start the server via Azure CLI. Every attempt fails with the following error: (ServerBusyWithOtherOperation) Cannot perform 'Restart' server operation because server is busy processing other operation.

I have run a deep investigation using the Observability Agent, which confirms that there is no workload saturation and no application-level errors. This clearly indicates an infrastructure-level provisioning stall.

Support concerns: When attempting to open a formal support ticket, the Azure portal directs me to purchase a paid support plan. However, since this is a clear infrastructure-level failure (as confirmed by the "Unknown" health state and the inability to process system commands), it is not appropriate to charge for technical support. This is a platform-side issue that prevents the service from functioning as expected.

Since I cannot resolve this via the portal or CLI, how can I force a reset of the provisioning process or get this instance back to a 'Ready' state without being forced into a paid support plan for a platform fault?

0 comments

Answer accepted by question author

Pilladi Padma Sai Manisha 11,715 Microsoft External Staff Moderator

Hi @Alessandro Vernile
Our backend team completed an investigation and identified that the server was stuck due to an internal platform-side provisioning issue, which prevented the service from completing the Starting operation. The engineering team performed the necessary backend recovery actions to clear the stalled operation and restore the service state.

The server has now been successfully recovered and is back in the Ready state. You should now be able to connect to the server and perform normal management operations (Start/Stop/Restart) without encountering the ServerBusyWithOtherOperation error.

If you continue to experience any issues, such as the server becoming stuck again or management operations failing, please let us know with the latest timestamp (UTC) and the server details so that we can investigate further.

Thank you for bringing this to our attention.

0 comments

1 additional answer

Your answer

Answer 1

Alex Burlachenko 25,285 MVP Volunteer Moderator

Alessandro Vernile hi, thx for sharing urs issue here at Q&A portal,

the server is stuck in a control-plane operation, not a PostgreSQL workload issue. ServerBusyWithOtherOperation means the PostgreSQL RP still thinks a start/stop operation is running. Once it’s stuck there, portal CLI retries usually won’t clear it bc every new command hits the same lock.

There isn’t a safe customer-side ‘force reset provisioning’ button for Flexible Server. If Resource Health is Unknown (Unplanned) and the server is stuck in Starting, this needs backend cleanup by Azure PostgreSQL support.

Grab the Activity Log entries for the original stop/start operation and the failed restart attempts. The correlation IDs matter more than screenshots here.

az monitor activity-log list \

--resource-id '<server-resource-id>' \

--offset 24h

https://learn.microsoft.com/en-us/azure/service-health/resource-health-overview

For support, use Help + support > Create a support request > Service and subscription limits / Service issue if available. Billing/account support is free, but technical support normally depends on plan. That said, for a platform health issue, MS still needs the resource ID, region, UTC time, and correlation IDs to route it.

My take is don’t keep sending restart/stop commands. It won’t unlock a stuck RP operation and can muddy the logs. Open a support case and ask them to clear/recover the stuck Flexible Server operation from the backend.

rgds,

Alex

&

If my answer was helpful pls mark it and additional thx if u follow me at Q&A portal

Alessandro Vernile 40 Reputation points

2026-06-25T08:01:03.78+00:00

Hi, thank you very much for the clarification and for confirming that this is a backend control-plane issue.

I have stopped all further manual restart/stop commands as suggested. I have attempted to follow your advice and open a support case via the 'Help + support' portal, but I am currently unable to do so: the portal wizard forces me towards the community support options and does not allow me to submit a technical request, even though I have an active support plan.

Since I cannot open a formal support ticket, I am sharing the logs here in the hope that they can be flagged for backend review:

- Timestamp: 2026-06-24T08:10:42Z

Could you please provide guidance on how to escalate this when the portal wizard prevents ticket submission for a platform-side issue? Thank you again for your professional guidance.
Alessandro Vernile 40 Reputation points

2026-07-07T07:35:21.7133333+00:00

Hello,

I am writing to provide an urgent update, as this issue has now persisted for two weeks without resolution.

Unfortunately, I have received no further response following our last communication via private message, and the server remains stuck in a "Starting" state. This situation is causing a severe disruption for my client, with their production environment being offline for 14 days.

I must stress a fundamental distinction: I am not requesting "technical assistance" for database configurations or optimizations. I am reporting a major infrastructure-level failure of the Azure control plane, which has locked the resource in an inconsistent state.

I have already attempted to contact support through the "Billing and Subscription" section (the only channel accessible without a paid support plan), but they refused to escalate the issue to the technical team, insisting that I upgrade my support plan. I find it unacceptable that a customer is forced to incur additional costs to restore an Azure service that is down due to a clear failure of Microsoft's own infrastructure.

As this is a platform-side service outage, I am requesting an immediate backend intervention to clear the stuck operation and restore the resource. Could you please clarify how I can obtain immediate attention from the engineering team, given that standard channels are blocked and the private chat has remained silent?

I look forward to an immediate resolution.
Erland Sommarskog 137.4K Reputation points MVP Volunteer Moderator

2026-07-07T08:29:06.5066667+00:00

I happened to see this, because you posted a new Question, which the spam filter on this site did not like. According to the title of that post, this is a production server.

I would have some sympathy if this was just a development instance that you use for your own personal training. In that case, it is understandable that you are not inclined to pay for a support contract. That is what I'm doing with my Azure subscriptions; I don't have any support contract. But obviously, if something goes wrong on the backend, you want Microsoft to sort this out. As for why nothing happened after you replied to Pilladi's private message, I leave to Pilladi to explain that. But as support staff has more issues to deal with, things may be forgotten. That's human. Also, since you don't have a support contract, you will be a lower priority.

But if this now is a production server, you should have a support contract in my opinion. Now, I don't know exactly what sort of production environment this is; maybe it is only for the local yachting club. Nevertheless, the cheapest support plan is only 30 USD/month. That is intended for developers, so it is not really feasible for production, but maybe it is sufficient for your needs. In worst case, you would have to sign up for a support plan and then cancel it, when this issue has been resolved. But if you want to avoid future outages, it may be better to pay that insurance of 30 USD per month.

(By the way, since that other question was a duplicate with your post above, I did not undelete it.)