bug#41625: Sporadic guix-offload crashes due to EOF errors

[Maxim Cournoyer]

Hello,

Maxim Cournoyer <maxim.cournoyer@gmail.com> writes:

> Hi Marius,
>
> Marius Bakke <marius@gnu.org> writes:
>
>> Maxim Cournoyer <maxim.cournoyer@gmail.com> skriver:
>>
>>>> Is running ‘guix offload test /etc/guix/machines.scm overdrive1’ on
>>>> berlin enough to reproduce the issue?  If so, we could monitor/strace
>>>> sshd on overdrive1 to get a better understanding of what’s going on.
>>>
>>> It's actually difficult to trigger it; it seems to happen mostly on the
>>> first try after a long time without connecting to the machine; on the
>>> 2nd and later tries, everything is smooth.  Waiting a few minutes is not
>>> enough to re-trigger the problem.
>>>
>>> I've managed to see the problem a few lucky times with:
>>>
>>> --8<---------------cut here---------------start------------->8---
>>> while true; do guix offload test /etc/guix/machines.scm overdrive1; done
>>> --8<---------------cut here---------------end--------------->8---
>>
>> I used to be able to reproduce it by inducing a high load on the target
>> machine and just let Guix keep trying to connect.  But now I did that,
>> and set overload threshold to 0.0 for good measure, and Guix has been
>> waiting patiently for two hours without failure.
>>
>> So AFAICT this bug has been fixed.  Perhaps Berlin or the Overdrive
>> simply needs to be updated?
>
> Ah!  Do you have root access to overdrive1?  It'd be interesting to
> reconfigure it to update the guix-daemon and see if the problem
> vanishes.

Good news, this seems resolved with the newer Guile-SSH 0.15.1, where
long delays to return some output no longer triggers an EOF response
(instead now the client waits still).  I believe it was fixed by this
commit [0].

Many thanks to Artyom Poptsov for fixing it!

Closing.

Maxim

[0]  
https://github.com/artyom-poptsov/guile-ssh/commit/fefaab9e925d015b01abc7c76ea4017c373ad895