Delete room (admin-api) fails/timeouts before done.

[olmari]

Description

When I use delete room through Synapse admin-api, no matter v1 or v2, if it is massive room, after some 4-6 hours (not really timed) or so the delete task will fail, despite there was being PSQL thread for delete for full load all that time... But if I just keep rerunning the task, eventually it will success.. Here is an excerpt from one room through v2 that I first tried couple times with v1 api, then with v2:

{
  "results": [
    {
      "delete_id": "UmVcAHOubayaZUVX",
      "status": "failed",
      "shutdown_room": null
    },
    {
      "delete_id": "eRURlGTxWZdZalLa",
      "status": "failed",
      "shutdown_room": null
    },
    {
      "delete_id": "vDZzfLohvBiHCJke",
      "status": "complete",
      "shutdown_room": {
        "kicked_users": [],
        "failed_to_kick_users": [],
        "local_aliases": [],
        "new_room_id": null
      }
    }
  ]
}

Steps to reproduce

• Have massive room to delete
• Hit /_synapse/admin/v2/rooms/<RoomID>
• Wait.

Homeserver

hacklab.fi

Synapse Version

1.116.0

Installation Method

Debian packages from packages.matrix.org

Database

Single PSQL 17

Workers

Multiple workers

Platform

Debian 12, 8 core, 32g RAM Upcloud VM, 100k IOPS-

Configuration

Presence is enabled, MSCs for Hookshot E2E are enabled.

Relevant log output

Haven't found the relevant spot, but as the thing _eventually_ works, I'm 99.9% sure it is some form of timeout.

Anything else that would be useful to know?

No response