How to Manage Azure PostgreSQL Flexible Servers Rejecting All Connections & Unresponsive Control Plane

Select Language:

Our development and staging environments are currently completely offline. Because of this, no applications that rely on these databases can function. We also can’t do any deployments or testing until the issue is fixed.

The problem involves two Azure Database for PostgreSQL Flexible Server instances. They are in the same subscription but in different resource groups, and both are showing the same symptoms at the same time. These servers were started this morning as part of our usual morning routine, but they have had connection issues since then.

From a user’s perspective, any attempt to connect is rejected with an error message that says:

“FATAL: pg_hba.conf rejects connection for host ““, user ““, database ““, SSL encryption.”

These connection errors happen whether you’re connecting from our office IP (which is on the allowlist), from AKS pods in the same subscription, or from Azure Cloud Shell. This shows that the issue isn’t just an IP list problem.

We checked the configuration on both servers, and they’re identical. No recent changes or updates have been made that could have caused this issue. The admin user account involved in these connections was working fine before today.

Even though the servers say they’re “Ready,” the control plane isn’t responding properly to management commands. For example, when trying to stop or restart the servers, the commands are rejected with a “ServerBusyWithOtherOperation” error. Moving to disable public access temporarily is accepted but then hangs for over an hour without doing anything. Creating or deleting firewall rules seems to work, but it doesn’t resolve the connection rejection problem, implying that the control plane isn’t syncing the pg_hba.conf file with the data plane.

We’ve performed several diagnostics:

– No resource locks are active.
– There are no read replicas or high-availability setups.
– Service health in the region is fine, with no reported incidents.
– Both servers are healthy and show no issues with CPU, memory, or storage.
– Their configurations are identical.
– The IP addresses don’t seem to matter, as even Cloud Shell connections are rejected.

This problem matches a known Microsoft issue (Q&A case 5732231). The symptom involves a “stuck” server where the pg_hba.conf file isn’t updating properly despite active firewall rules, which points to a control plane and data plane sync failure. As a workaround, Microsoft recommends toggling public network access, but we’ve already tried this, and now the servers are hung at that step with no resolution.

At this point, debugging further is very difficult since the servers reject all interactions and appear completely frozen. We’re seeking guidance on how to resolve this issue and restore functionality to our environments.