Failed to send_join via any server for #emacs:matrix.org

Description

Joining #emacs:matrix.org fails with a Failed to send_join via any server. Join method was the room directory via element web.

Steps to reproduce

• Search emacs in the matrix.org room dir via ele web
• Click and join the room

2021-05-19 16:19:31,049 - synapse.federation.federation_client - 548 - WARNING - POST-1001498 - Failed to send_join via matrix.org
Traceback (most recent call last):
  File "/opt/synapse/synapse_git/synapse/federation/federation_client.py", line 517, in _try_destination_list
    return await callback(destination)
  File "/opt/synapse/synapse_git/synapse/federation/federation_client.py", line 711, in send_request
    room_version=room_version,
  File "/opt/synapse/synapse_git/synapse/federation/federation_client.py", line 427, in _check_sigs_and_hash_and_fetch
    ).addErrback(unwrapFirstError)
  File "/opt/synapse/env/lib/python3.7/site-packages/twisted/internet/defer.py", line 761, in send
    raise result.value
twisted.internet.defer.FirstError: FirstError[#0, [Failure instance: Traceback: <class 'synapse.api.errors.SynapseError'>: 403: event id $1449445060123025NBoor:matrix.org: unable to verify signature for sender ghostcorp.net: 400: Not signed by ghostcorp.net
/opt/synapse/synapse_git/synapse/federation/federation_base.py:76:_check_sigs_and_hashes
/opt/synapse/synapse_git/synapse/federation/federation_base.py:219:_check_sigs_on_pdus
/opt/synapse/env/lib/python3.7/site-packages/twisted/internet/defer.py:348:addErrback
/opt/synapse/env/lib/python3.7/site-packages/twisted/internet/defer.py:330:addCallbacks
--- <exception caught here> ---
/opt/synapse/env/lib/python3.7/site-packages/twisted/internet/defer.py:662:_runCallbacks
/opt/synapse/synapse_git/synapse/federation/federation_base.py:216:sender_err
]]

Which seems to suggest that room is broken from the perspective of my HS?

This has the side effect that my voyager bot tries to join the room over and over, causes a membership event but never actually joins from my synapse perspective.

Perspective from other servers is that synapse sends membership join events over and over again. On Element android this spams the timeline and as my synapse doesnt realise this I cannot really prevent this in my code I think unless I cache this data myself.

Version information

• Homeserver: matrix.nordgedanken.dev

If not matrix.org:

• Version: 1.34.0 (b=develop,9752849e2,dirty)

• Install method: Git clone with pip

• Platform: Debian 10

[richvdh]

This issue seems to have become a bit of a pile-on for everyone who ever saw any sort of error joining any sort of room. I was able to join #emacs:matrix.org with no problem, so I think the original problem has probably been solved, and I’m going to close this.

If people have problems joining rooms, please create a new issue rather than piling onto one that sounds vaguely related - there are a lot of reasons a join can fail and it’s much easier for us to close duplicate issues than it is to split out unrelated problems.

[erikjohnston]

That stack trace is really odd, we should be catching that error correctly:

https://github.com/matrix-org/synapse/blob/1350b053da45c94722cd8acf9cfd367db787259c/synapse/federation/federation_client.py#L391-L394