tunnel doesn't open / don't accept new connections in some unknown circumstances

[converge]

Hi team!

We at Kubeifrst have been using ngrok-go quite a while, it works pretty well. the only issue we have is that 1 in ~50 times we create a tunnel, the tunnel fails, and we're having a hard time to identify the root cause.

We had a previously memory leak (we didnt close the channel and connection instances), but we fixed the memory leak issue in this PR[1].

Is there any possibility that you guys could have a look, and give us some direction of what could be causing the issue?

p.s: we weren't able to reproduce the issue after this PR (so far). We don't use a token(token is empty). related report issue: https://github.com/kubefirst/kubefirst/issues/939

[1] https://github.com/kubefirst/kubefirst/pull/1228

[russorat]

Hi @converge, thanks for the issue! Looking through the linked issue, is it possible you were testing us during an unexpected service outage? I'm not sure if the timestamps line up. https://status.ngrok.com/incidents/3y0t4x2zpmz3

[converge]

@russorat tks for the prompt reply!

Following our issue history, we started to identify the issue 15 of december, and since them, we have been tracking and trying to figure out what could be issue. Last week we found a pattern we could reproduce for (maybe 2 Kubefirst installations[1]), but we could not reproduce that again later on.

I have been thinking to check the request token payload, and look for this condition[2], if the error occurs, close the tunnel and try to open a new, wdyt?

[1] Kubefirst opens a ngrok tunnel during the installation, and closes it at the end(when user exists main thread) [2]