PostgreSQL Flexible Server: management operations fail (restart, database list) since platform maintenance; server stays Ready and serves clients

ZakeyF-3559 0 Reputation points
2026-11-13T19:09:28.0733333+00:00

Problem description

Since a platform maintenance restart on 2026-11-07, management operations on one of our Azure Database for PostgreSQL flexible servers have failed. The server reports Ready and keeps serving client connections normally, but:

  • A restart requested at 2026-11-10 16:31:51 UTC was accepted, then failed at 16:32:52 UTC with ResourceOperationFailure / InternalServerError (tracking ID 603bf414-9296-47f9-bccf-3d2e21670456). The server did not restart.
  • Listing databases still fails, as of 2026-11-13 17:26 UTC:
az postgres flexible-server db list --server-name <server> -o json
ERROR: (ServerUnableToRetriveData) Server encountered an issue while trying to retrieve some data. Try again later.
Code: ServerUnableToRetriveData

Support case 2611100040005785 is open for this.

Environment

  • Azure Database for PostgreSQL flexible server, PostgreSQL 16
  • Burstable, Standard_B2ms (2 vCores, 8 GiB)
  • Region: Central US
  • Private access (VNet-integrated), public network access disabled
  • No configuration changes since a successful deployment on 2026-11-06

What I've already tried

  • Restart: one attempt on 2026-11-10, which failed as above. Support advised against retrying, so I haven't.
  • Checked the server state: az postgres flexible-server show returns state: Ready, sku: Standard_B2ms, tier: Burstable.
  • Checked Azure Monitor for resource pressure. Support's diagnostics report 10 out-of-memory events between 2026-11-07 17:07 and 2026-11-08 17:00 UTC, but the server's own metrics don't show memory pressure:
    • hourly maximum memory_percent was 24–31% during that window, and hasn't gone above 47% since 2026-11-06;
    • cpu_credits_remaining stayed at 911–917, so the Burstable limit was never hit.
  • Looked at what changed on 2026-11-07. Right after the scheduled maintenance restart (around 08:00–09:00 UTC):
    • the active_connections metric stopped reporting and is still missing;
    • cpu_credits_remaining froze at exactly 911;
    • Resource Health logged about 27 "Unavailable – PlatformInitiated – Unknown reason" events;
    • baseline CPU roughly doubled (hourly maximums about 11–13% before, 16–28% after) with no change in our traffic.
  • Held off on a compute change. Support suggested moving to General Purpose to land on fresh compute. I haven't done it yet, because their note also says the change can fail while management operations are stuck. A General Purpose server of the same size also has 8 GiB of memory.

Current status

  • The server is Ready and serving clients normally.
  • Restart and database listing still fail.
  • The active_connections metric is still missing.

Questions

  1. Can the platform management agent on this server be repaired, or the server moved to a new host, from Microsoft's side?
  2. Is a compute or SKU change likely to succeed in this state, or would it fail with the same error and leave the server worse off?
  3. Which process produced the 10 out-of-memory events?
  4. Are automated backups and point-in-time restore still working while management operations fail?

I can share the server name and subscription details privately with a Microsoft engineer.

Azure Database for PostgreSQL
0 comments No comments

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.