Edit this page | Blame

Updating Guix profiles (for production and development)

This document describes updating Guix for GeneNetwork by using a channel. To develop guix packages see

Executive summary

To build guix profiles that can run GN2+GN3 we use a so-called channel.

Note that a recently tested channel can be fetched from cd.genenetwork.org. That should include the recent git hash values.

Alternatively put the following into a channels.scm file.

(list (channel
       (name 'gn-bioinformatics)
       (url "https://gitlab.com/genenetwork/guix-bioinformatics")
       ;; (url "https://git.genenetwork.org/guix-bioinformatics/guix-bioinformatics")
       (branch "master")))

Build a profile using

$ guix pull -C channels.scm -p ~/.guix-extra-profiles/genenetwork

Activate the newly built profile using

$ export GUIX_PROFILE=~/.guix-extra-profiles/genenetwork
$ . $GUIX_PROFILE/etc/profile

Hack away on any GeneNetwork repository as usual. Now, any development environment you create with `guix shell` will be identical to what others on the team and the continuous integration system are using. This development environment is guaranteed to not break regardless of the latest changes to Guix upstream or other externalities.

Production (tux01)

After downloading channels.scm from

On production we do something like

tux01:~$ ~/opt/guix-pull/bin/guix pull -C channels.scm -p ~/opt/guix-gn-channel
tux01:~$ ~/opt/guix-gn-channel/bin/guix package -i genenetwork2 -p /usr/local/guix-profiles/gn-latest-yyyymmdd

Now we can use that profile to run genenetwork2 and genenetwork3:

source /usr/local/guix-profiles/gn-latest-yyyymmdd/profile

And everything should be in the PATH, PYTHONPATH etc.

Rationale

From time to time, updates to upstream Guix break our guix-bioinformatics channel. As a result, `guix pull` breaks and our work comes to a standstill for a few days until this can be fixed. While it is important to gradually move towards the latest and greatest, we would like to avoid being at the mercy of externalities and would prefer to update in a more controlled way without interrupting everyone's work.

To this end, we hard-code the guix-bioinformatics channel to depend on a *specific* commit of upstream Guix that is tied to guix-bioinformatics, for example:

(list (channel
        (name 'gn-bioinformatics)
        (url "https://gitlab.com/genenetwork/guix-bioinformatics")
        (branch "master")
        (commit
          "feff05b47c305d8c944499fbc00fd2126f2b881d")))

This is why the recommended channels.scm file above does not include a %default-guix-channel. However, this comes with the drawback that your entire system will be stuck at that specific commit of upstream Guix (but not if you use another profile as described above). We highly recommend using a separate `guix pull` profile specifically for GeneNetwork work, as described above.

This scheme also comes with the added bonus that all members on the team and the continuous integration system will be using exactly the same Guix. Above channels.scm file is only exposed on a *succesful* build.

Notes

We recently had to switch to gitlab because our git server went down on Penguin2. We may move to a cgit solution soon, see


(made with skribilo)