This repository has been archived by the owner on Apr 26, 2024. It is now read-only.
Synapse v1.50.0 starts and federation tester reports as good but clients cannot connect #11762
Labels
S-Critical
Blocks development, potential data loss, more than 25% of users possibly affected, no workarounds.
T-Defect
Bugs, crashes, hangs, security vulnerabilities, or other reported issues.
Comments
scottwallacesh
changed the title
|
I have exactly the same problem |
|
Hi @scottwallacesh! Thank you for making an issue. A few questions to try and understand the problem:
Thank you! |
|
|
Interesting, thank you @jonasled. Testing locally using the 1.50.0 docker image with a default config, I seem to receiving Could you paste the |
|
I'm also curious about how you're deploying. E.g., @jonasled are you also using Docker? Which image are you using: the one we publish as |
|
I've attached the listener section below: |
callahad
added
P1
T-Defect
|
Thanks @jonasled. Note to others that copying those settings exactly on my local setup seems to not produce any issues. |
|
Interesting. I use a postgres server for the database, have added multiple bots and in front of the synapse is as reverse proxy a traefik 2. Downgrading the docker to version 1.49.0 resolved the issue. |
|
@jonasled I assume your Postgres server version is at least 10+? Note that Synapse v1.50.0 dropped support for older versions. @scottwallacesh would you happen to also be using traefik as your reverse proxy? |
|
Postgres is version 13 |
|
Same problem here (Debian bullseye container, postgres 14, running behind an apache proxy) |
|
@anoadragon453 I am using Traefik, yes. I think this might be related to CORS headers but I'm finding it hard to work out why that's changed in this release. Reverting to v1.49.2 gets everything working again. |
|
I've uploaded my complete config (without keys & passwords) here if it helps: https://paste.jonasled.de/riguvifeqi.yaml |
|
No problems with 1.50.0 behind Traefik and Postgres 13 |
|
Thanks @scottwallacesh. Are you seeing the 404's appear in your Synapse logs as well? If this were a CORS issue then the client would refuse to make the request, and thus it would never reach your server. |
|
I have a strange issue as well, I have 6 servers on 4 domains & 2 subdomains, all on the same host docker. All servers are identical in config but only one subdomain & one domain are working, the rest cant connect to. From the only domain that is working, I can access the others via federation once logged in on one of the working servers. found out the 2 working server had no |
|
These are the requests I saw in the debugging tools |
|
We've managed to narrow this down to having the |
|
Removing the webclient listener works |
No 404s in my Synapse logs, no. The reason I suspect CORS header is because I see lots of CORS issues in the brower console and when making a HTTP request to Synapse I see CORS headers in v1.49.2 and none in v1.50. v1.49.2: v1.50.0: |
|
Note that #11763 has been created with more details on the issue and will serve to track the fix. |
Yep ... this fixed it for me. |
|
A fix for this issue was implemented in #11764 and has been released in Synapse v1.50.1. Thank you all for your swift help in debugging this ❤️ |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Description
Synapse v1.50 starts correctly and the federation tester shows everything as working but Android and Web clients cannot connect.
Steps to reproduce
Element Web says the server is offline and unavailable.
Element Android connects but when posting a message show the 'Retry' button(s).
Version information
Homeserver: Synapse
Version: v1.50
Install method: Docker container (
matrixdotorg/synapse:latest)Platform: Linux - Docker CE
The text was updated successfully, but these errors were encountered: