unexpected EOF after 15 minutes

[kgosciniak-insys]

@chmouel

When running gosmee in client mode after about 15 minutes the application crashes with error:

Wed, 10 Apr 2024 07:09:00 UTC INF Starting gosmee version: dev

Wed, 10 Apr 2024 07:09:01 UTC INF 2024-04-10T07.09.04.177 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:09:31 UTC INF 2024-04-10T07.09.04.826 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:10:02 UTC INF 2024-04-10T07.10.04.594 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:10:18 UTC INF 2024-04-10T07.10.04.009 pullrequest event replayed to https://another.example.org, status: 200
Wed, 10 Apr 2024 07:10:22 UTC INF 2024-04-10T07.10.04.294 pullrequest event replayed to https://another.example.org, status: 200
Wed, 10 Apr 2024 07:10:52 UTC INF 2024-04-10T07.10.04.711 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:11:24 UTC INF 2024-04-10T07.11.04.898 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:11:57 UTC INF 2024-04-10T07.11.04.808 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:12:32 UTC INF 2024-04-10T07.12.04.056 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:13:09 UTC INF 2024-04-10T07.13.04.668 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:13:49 UTC INF 2024-04-10T07.13.04.211 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:14:30 UTC INF 2024-04-10T07.14.04.993 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:15:16 UTC INF 2024-04-10T07.15.04.981 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:16:04 UTC INF 2024-04-10T07.16.04.320 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:17:32 UTC INF 2024-04-10T07.17.04.748 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:18:46 UTC INF 2024-04-10T07.18.04.419 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:20:25 UTC INF 2024-04-10T07.20.04.181 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:22:01 UTC INF 2024-04-10T07.22.04.048 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:23:18 UTC INF 2024-04-10T07.23.04.374 Forwarding https://example.org to https://another.example.org
Wed, 10 Apr 2024 07:24:49 UTC INF 2024-04-10T07.24.04.215 Forwarding https://example.org to https://another.example.org
2024/04/10 07:25:19 unexpected EOF

docker ghcr.io/chmouel/gosmee client https://example.org https://another.example.org

[chmouel]

i think it's best running over caddy to ensure long polling works...

[kgosciniak-insys]

@chmouel
Correct me if I'm wrong but there is supposed to be one long connection between client and server right?
If I see multiple Forwarding X to Y logs that means there is reconnection and this is incorrectly configured?

[ni-skopp]

Hi @chmouel by using Caddy you mean run reverse proxy in front of gosmee server? We have same issue right now, running gosmee server in kubernetes cluster with Loadbalancer in front and gosmee client as binary on linux machine.

[kgosciniak-insys]

I've managed to bypass this problem by running very long timeout between load balancer and gosmee server as described here:
https://github.com/chmouel/gosmee/?tab=readme-ov-file#nginx

There is also requirement to only have only one instance of gosmee server active, because client doens't work with multiple instances (missing redis integration from smee.io)

[chmouel]

There is also requirement to only have only one instance of gosmee server active, because client doens't work with multiple instances (missing redis integration from smee.io)

can you describe me your use case ? i usually have multiple client running for different endpoint listening.

[kgosciniak-insys]

There is also requirement to only have only one instance of gosmee server active, because client doens't work with multiple instances (missing redis integration from smee.io)

can you describe me your use case ? i usually have multiple client running for different endpoint listening.

Sure, many organization are required to limit single point of failures in their infrastructure. Since you are allowed to run only one instance of gosmee server because there is no synchronization between servers so when client connects through load balancer to instance B and webhook goes to instance A, client wouldn't get notification. Original smee.io had this resolved by using Redis:
https://github.com/probot/smee.io?tab=readme-ov-file#running-multiple-instances-of-smeeio

[chmouel]

ah right thanks for pointing this out, that's an interesting thing to implement, but since i only work on gosmee on a limited free time i can't promess a timeline when this will be implemented..