[Bio-packaging] Packaging

Pjotr Prins pjotr.public66 at thebird.nl
Mon Jun 8 10:40:51 UTC 2015


On Mon, Jun 08, 2015 at 06:22:11AM -0400, Brad Chapman wrote:
> I'm still leaning on Docker as the solution to a lot of the
> underlying issues. I've been patiently waiting for user namespace
> support which will allow users to run Docker containers without
> needing root-equivalent privileges. When that lands I'm hoping to
> push it more as a solution. If that takes off then we can do any
> sort of package management we want inside the container and will not
> need to worry about root installation.

When you have Docker it should be possible to have it play with Guix.
Docker effectively gives you root access and a root folder (if I am
correct and permissions allowed). I don't use Docker, but it would be
nice if someone managed to get that ball rolling.

But yes, we are going to make do for a while. I think the future for
bioinformatics will be VMs with or without Docker. It is kinda ironic
you were way ahead with CloudBiolinux years ago! We have to push
harder against the current state of affairs where we can't dictate
what we install on clusters. GNU Guix may actually help - with or
without VM - since there are no real security concerns and packages
are well isolated which means you can't mess up the main system.

Give it another 5 years and half the clusters may allow Guix and/or
VMs. 

> Hope this helps with my current practical needs. I appreciate all
> the work on Guix and improving package management, Brad

We'll keep trying. I have root on the systems that matter for
genenetwork.org. So GNU Guix has gone in :). Arvados will get Guix
support too. I have not written many packages yet, but that will come.

By the way the Arvados BWA-GATK pipeline pilot was a great success.
Performance is flat for 8 parallel jobs (8x70GB human fastq variants
called in 1.5 days), i.e.  perfect scalability.  I need to write a
BLOG entry about that.

Pj.

-- 


More information about the bio-packaging mailing list