Notification server seems to not work properly about seafile-containerized HOT 15 CLOSED

Hi,

how are you testing the functionality of the notification server? I only tested the ping endpoint and assumed it was working. The client does not show whether it is using the notification server, so it is hard to retrace if it is functional.

from seafile-containerized.

I am using https://mitmproxy.org/ to proxy calls from windows desktop client to the seafile instance.

Steps:

1. download mitmproxy
2. run mitmproxy --mode upstream:machineip:8068 where machineip is my local network seafile instance (homeserver)
3. change seafile client server address to http://localhost:8080/ or whichever port mitmproxy binds to (shown bottom right in the terminal), the desktop client must run on the same machine as mitmproxy
4. see that /ping does return 200, but /notification returns error 400
5. see that desktop client is retrying the connection, unsuccessfully every time

from seafile-containerized.

Ok, but you are just sending a GET request without any parameters. The server might expect another type of request, like POST and some parameters.

from seafile-containerized.

I am not sending any request, the request is comming from the desktop client, through the mitmproxy, what you see it what the official app sends, as I have stated. The desktop client is this:

from seafile-containerized.

I see...

I just checked on my setup and the client logs:

[07/26/23 20:42:20] Notification server is enabled on the remote server **URL**

The path is:
C:\Users\$USERNAME\ccnet\logs\seafile.log

Can you see the same?

According to the documentation, this means that the notification server works.

from seafile-containerized.

Yes, I can see that. But it logs at the time it sends the ping command, not the actual /notification

Websockets open once, then they no longer reconnect if everything works correctly. The ever repeating error 400 suggests that the connection cannot seem to be established.

It might support, but be blocked by caddy

from seafile-containerized.

I also tried to debug it with mitmproxy but I can see just a request to the ping endpoint. No other requests to any notification-related endpoint.

from seafile-containerized.

You mean, you do not see the /notification endpoint requests that I posted?

from seafile-containerized.

Exactly, I start the client and see one /notification/ping request and that's it. Not a single /notification request.

Regarding WebSocket, Caddy apparently supports upgrading connections to WebSocket by default without any additional config. So I doubt that the error lies there.

from seafile-containerized.

Would it be possible to post your whole seafile.conf ?

from seafile-containerized.

Here you go:

[fileserver]
port = 8082
max_upload_size=100000
max_download_dir_size=100000
use_go_fileserver = false

[database]
type = mysql
host = db
port = 3306
user = seafile
password = ...
db_name = seafile_db
connection_charset = utf8

[notification]
enabled = true
host = 0.0.0.0
port = 8083
log_level = debug
jwt_private_key = ...

from seafile-containerized.

except for the max_upload_size and max_download_dir_size, the only difference is that I have use_go_fileserver = true. When I changed it to false nothing changed. I am still getting the failed request to /notification

I dont have any blocked ports in my local network. Well, this is weird.

Do you have current seafile client version - 9.0.3 ?

from seafile-containerized.

In your screenshot, the error 400 seems to occur exactly every 60 seconds. Maybe Nginx is terminating the WebSocket session every 60 seconds?

Yes, I'm also running 9.0.3

from seafile-containerized.

I connected directly to the seafile caddy, skipping nginx and I have the same issue. The 60s interval seems to be just a reconnect behavior

from seafile-containerized.

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

from seafile-containerized.