Re: guixsd install questions

[Ludovic Courtès]

Hello!

Divan Santana <address@hidden> skribis:

> Then do the install with this guile code:
>
> #+BEGIN_SRC scheme
>   ;; two devices in raid0 striped with LUKS full disk encryption.
>   (bootloader (grub-configuration (device "/dev/vdb")))
>   (mapped-devices (list
>                    (mapped-device
>                     (source (list "/dev/vdb1" "/dev/vdc1"))
>                     (target "/dev/md0")
>                     (type raid-device-mapping))
>                    (mapped-device
>                     (source (uuid "fb29c6f6-b2c0-4c87-8651-4962b7125dc0"))
>                     (target "crypt")
>                     (type luks-device-mapping))))
> #+END_SRC
>
>
> And this too:
>
> #+BEGIN_SRC scheme
>   (file-systems (cons (file-system
>                        (device "root")
>                        (title 'label)
>                        (mount-point "/")
>                        (type "ext4"))
>                       %base-file-systems))
> #+END_SRC
>
>
> The above fails. So tried another install with device like so

Do you know how it fails?

My guess is that you’d need to explicitly mark one of the mapped device
as depending on the other; this cannot be guessed.

If you run “guix system shepherd-graph” on your config you’ll probably
see that there’s no such dependency.

Currently dependencies among mapped devices cannot be expressed, but
that’s easy to fix (by providing a ‘dependencies’ field as in
‘file-system’.)

> That failed, I then tried the UUID method, via =blkid
> /dev/mapper/crypt=, get the UUID and did another install with this
> snippet instead:
>
> #+BEGIN_SRC scheme
>   (file-systems (cons (file-system
>                        (device (uuid "4dab5feb-d176-45de-b287-9b0a6e4c01cb"))
>                        (title 'uuid)
>                        (mount-point "/")
>                        (type "ext4"))
>                       %base-file-systems))
> #+END_SRC
>
> This fails with waiting for root device.

Probably same problem as above.

For now, using just RAID or just LUKS will work; it’s the combination
that’s not supported yet.

> * How to recover a failed install. How to chroot a broken system and
> fix?
>
> You can see why I'm asking this. When my failed crypt install fails, I
> sometimes just want to reconfigure the system to try another method.
>
> Now when I run =guix system init /mnt/etc/config.scm /mnt= to recover
> the install to the same preveiously install disk it re-downloads,
> re-compiles and redoes the whole install, instead of just perhaps
> changing grub to (attempt to) fix my crypt issue.
>
> Ideally I want to chroot into the installed (and broken) environment and
> do a =guix system reconfigure /etc/config.scm=.
>
> How can one do this?

I guess you could boot the install image, mount the target file system,
chroot in it, run guix-daemon in there, and run ‘guix system
reconfigure’ there.

That should work though that’s inconvenient at best.

> * How to use a proxy to do the install
>
> This is from the boot install media.
>
> I've read the docs on using proxy though it's not working like I expect.
> Prob doing something wrong.
>
> I've done the following
>
> On tt1 I did =herd stop guix-daemon=
> Then exported proxy like so:
> export http_proxy=http://server.domain.co.za:8080/ ; export 
> ftp_proxy=$http_proxy ; export https_proxy=$http_proxy
>
> =herd start guix-daemon=

The ‘http_proxy’ variable needs to be set in the environment of the
‘guix-daemon’ process itself, which is why this doesn’t work.  We should
make it easier to choose a proxy, for instance by having the daemon
honor client-provided proxy settings.

In addition, note that ‘https_proxy’ and ‘ftp_proxy’ are not supported
yet.

Thanks a lot for your detailed feedback!

Ludo’.