bug-guix
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

bug#58290: guile ssh error on guix deploy


From: Ludovic Courtès
Subject: bug#58290: guile ssh error on guix deploy
Date: Tue, 18 Oct 2022 17:31:42 +0200
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/28.1 (gnu/linux)

Hi,

Marius Bakke <marius@gnu.org> skribis:

> I have the same problem.  Log messages around failure look something
> like this (extracted from above message):
>
>   Oct  4 11:51:49 localhost shepherd[1]: Service sshd-2 has been started. 
>   Oct  4 11:51:50 localhost sshd[250]: Accepted publickey for bob from 
> 185.70.53.164 port 1915 ssh2: RSA 
> SHA256:/X3jyhyMizAOKOjCfXK5cLN3Pa0vmi7dbQG+fxK3d3Y
>   Oct  4 11:52:28 localhost sshd[252]: error: no more sessions
>   Oct  4 11:52:28 localhost shepherd[1]: 1 connection still in use after 
> sshd-2 termination. 
>   Oct  4 11:52:28 localhost shepherd[1]: Service sshd-2 has been disabled. 
>
> Then deploy crashes with 'Channel opening failure'.

“no more sessions” comes from this:

--8<---------------cut here---------------start------------->8---
int
session_open(Authctxt *authctxt, int chanid)
{
        Session *s = session_new();
        debug("session_open: channel %d", chanid);
        if (s == NULL) {
                error("no more sessions");
                return 0;
        }

[...]

}
--8<---------------cut here---------------end--------------->8---

Would be interesting to run sshd in verbose/debug mode and see why we
hit that; it could be because the maximum number of sessions has been
reached or something.

> Due to the number of SSH connections made by deploy and frequent
> occurence of this problem, I was not able to successfully deploy through
> many attempts.

Ouch.  Normally, thanks to memoization, ‘guix deploy’ opens only one
session per machine.  (I think it works well with the 25 local build
machines behind berlin; it’s also fine for a small set of machines I
take care of at work.)

> I found that removing the memoizing open-machine-ssh-session* in
> (gnu machine ssh) works around it and can happily deploy again.
>
> (that is, just use 'open-machine-ssh-session' instead)

Interesting.  That gives us a hint.  Could you add a ‘pk’ in
‘open-machine-ssh-session*’ to see how many connections it opens?

Thanks for your feedback!

Ludo’.





reply via email to

[Prev in Thread] Current Thread [Next in Thread]