Server Admin can't re-join room

[KubaChoice]

Description

If server admin accidentally leve (probably also kicked) from the space, there is no way how to get back to space (only direct DB manipulation).

This is very frustrating.
There is use case, that some external system wants synchronize users / spaces to correspond this system universe. But if this happen, synchronisation breaks because if this admin is not in space, can't manage users in it, and even ADMIN can't get back (not federated use-case but local ones).

I tried everything - probably BUG or is not possible at all (which is, IMO, stupid).

Steps to reproduce

AS admin:

• create space
• leave space
• try to get back

Homeserver

local one

Synapse Version

1.139.1

Installation Method

Docker (matrixdotorg/synapse)

Database

sqlite and postgres

Workers

Single process

Platform

Dockerized on Debian GNU/Linux 12 (bookworm)

Configuration

No response

Relevant log output

matrix_synapse  | 2025-11-14 08:39:28,492 - synapse.http.server - 135 - INFO - POST-9680 - <XForwardedForRequest at 0x7f428d180bf0 method='POST' uri='/_synapse/admin/v1/join/!vXvw...CENSORED' clientproto='HTTP/1.1' site='8008'> SynapseError: 403 - @admin:server not in room !vXv1Q...CENSORED.

Anything else that would be useful to know?

No response

[devonh]

This is a known issue with the matrix specification itself. (ie. matrix-org/matrix-spec#95)

To maintain spec compliance, there is nothing Synapse can do to alleviate the issue until the spec has been updated with a solution.
The only option right now would be to do as you mentioned and manually manipulate the database as required.

Closing this issue in favour of the matrix-spec issue.

[devonh]

I guess we can leave this issue open for now since there is no similar issue in Syanpse already.
But know that there is no resolution until the spec issue has been resolved.

There is #1533 which is similar, but that issue is specifically about the error returned.