Ricardo Wurmus <rekado@elephly.net> writes:

> Christopher Baines <mail@cbaines.net> writes:
>
>> Coupled with this, I think there are some services (hpc.guix.info, ...)
>> it would be nice to move off bayfront (a machine in Bordeaux) to
>> somewhere that's more reliable/remotely accessible, mostly to avoid the
>> risk of me breaking them again when attempting to reconfigure bayfront.
>
> That’s unfortunate.  Why is bayfront less reliable than we want it to
> be?  Surely that’s not because it’s using Guix System, because
> {berlin,ci}.guix.gnu.org also runs Guix System and has been reliably
> hosting services for us.

I guess having access to recover from issues is the primary concern, and
reliability relates to that.

Around FOSDEM, I think I caused Bayfront to restart while attempting to
restart cuirass. I'm not sure exactly what happened, as my SSH
connection just dropped, and the machine became inaccessible (including
Cuirass, hpc.guix.info, ...).

Luckily, some time later Ludovic and Andreas were available to remotely
access the power supply, and turn the machine off and on, this did seem
to bring the machine back.

> I would not mind hosting services on one of the 30 new servers the MDC
> bought for ci.guix.gnu.org, but I’d be happier if we figured out what’s
> wrong with bayfront instead of centralizing all services in Berlin…

I'm definitely not saying anything is wrong with Bayfront, just that it
would be good to maybe check what the expectations are with the services
it hosts. Given that the remote access only allows power cycling the
machine, if that doesn't work, there's the potential for downtime for
those services to last until someone physically goes to the machine to
fix it, if SSH access can't be restored by turning it off and on again.