Postgress Flexible server instance suddenly showing as "in recovery Mode"

Question

Postgress Flexible server instance suddenly showing as "in recovery Mode"

Anil Kumar 0

Postgress Flexible server instance suddenly showing all the databases showing in "read-only" mode and instance is showing in "recovery mode". It was deployed with Zone redundant HA.

We are connecting it using hostname/FQDN only. Already tried stop / restart the instance, is there any other things needs to be changed / reconfigured?

1 answer

Your answer

Answer 1

Pilladi Padma Sai Manisha 1,975 Microsoft External Staff Moderator

Hi Anil Kumar,
Your PostgreSQL Flexible Server entering "recovery mode" with read-only databases is a standard indicator of an ongoing high availability (HA) failover process, especially since it's configured with Zone-Redundant HA the standby replica is promoting to primary after detecting a primary issue.

Stop/restarts won't help as they extend the transition; monitor for auto-completion instead.

Check Current Status

In Azure Portal, go to Overview status should shift from "Recovery" to "Available" (typically 5-15 mins).
Under High availability blade, note Primary/Standby roles (they swap post-failover).
Resource Health (Support + troubleshooting) shows HA events like "Unplanned Failover".

Resolve if Stuck

Connect read-only and run: SELECT pg_is_in_recovery(); (returns true during recovery).

Trigger Planned failover from HA blade for quick resolution (~60s downtime).
Rule out storage full: Compute + storage >95% triggers read-only.

Monitor & Prevent

Set alerts on Monitoring > Alerts for Resource Health/HA metrics (e.g., is_read_only). Connection via FQDN remains valid no changes needed.

Share a screenshot of Resource Health/HA blade or server name for more targeted help. This resolves 99% of cases without data loss.

References:

HA Monitoring

Configure HA

Anil Kumar 0 Reputation points

2026-01-12T05:51:22.14+00:00

Hi @Pilladi Padma Sai Manisha , we have already checked health blade, it's showing green only.

In Azure Portal, overview status is showing "Available".

Resource Health (Support + troubleshooting) is not showing any "Unplanned Failover".

This DB instance was deployed using Virtual network Integration. A private DNS zone contains DB instance FQDN "A" record set, was mapped with Virtual network.

To resolve the DB instance FQDN, local host entry was configured on the source system pointing to 10.0.0.4 IP address.

Now after update the local host entry to 10.0.0.5, we are able to connect to Primary Instance.

Wondering why there is no notification in case any failover happened in the backend.
Pilladi Padma Sai Manisha 1,975 Reputation points Microsoft External Staff Moderator

2026-01-13T01:17:48.9433333+00:00

Hi Anil Kumar,
Thanks for confirming the blades are green and server shows "Available." That's expected after a Zone-Redundant HA failover the PostgreSQL server keeps the same FQDN, but Azure updates the Private DNS zone "A" record to point to the new primary's IP (10.0.0.5). Your local hosts entry pinned to 10.0.0.4 bypassed this, hitting the old standby until you updated it.

To prevent this going forward: Remove the hosts-file override completely and rely on Private DNS resolution for the FQDN (it auto-tracks the active primary). If using custom DNS in your VNet, forward queries to 168.63.129.16. Also double-check no resource locks on the Private DNS zone/record set they can block Azure's DNS updates during failover.
Manoj Kumar Boyini 4,000 Reputation points Microsoft External Staff Moderator

2026-01-14T18:15:07.86+00:00

Hi Anil Kumar,

I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.

Share via

Postgress Flexible server instance suddenly showing as "in recovery Mode"

1 answer

Your answer