[R-repo-dev] R/rpm part 2

[Dan Bolser]

Hi P,

I've CC'ed this reply to the new mailing list
(r-repo-dev at bioinformatics.org) to get things going there. Sorry if
this results in two copies of the mail arriving in your inbox. We may
as well move this kind of project discussion to the ML ASAP.

I'll reply to specific points 'inline' below:


On 23 April 2010 19:15, Pierre-Yves <pingou at pingoured.fr> wrote:
>
> Dear all,
>
> There is the beginning of part 2.
>
> We have a group of people ready to go so let's get further.
>
> Allen has come with a nice piece of code to generate the list of RPMs
> which can be generated without dependencies, then with the dependencies
> of the first iteration, then of the second and so on.
> The code (written in R) is available there:
> http://pingoured.fr/bioc/calc_buildorders.R
> I have adapted it so that it outputs the first iteration to a file in
> the current directory at the end of the process.
>
> While running it within R you will have an output quite verbose but the
> end will be similar to this:
>  Iteration 1; 2036 packages remain.
>  Iteration 2; 1296 packages remain.
>  Iteration 3; 817 packages remain.
>  Iteration 4; 371 packages remain.
>  Iteration 5; 108 packages remain.
>  Iteration 6; 36 packages remain.
>  Iteration 7; 8 packages remain.
>  Iteration 8; 0 packages remain.

Great. Can you upload the output to the project FTP site (or similar)?

http://www.bioinformatics.org/project/?group_id=924
http://ftp.bioinformatics.org/pub/r-repo/


> I have therefore compared this list to what I already had built and
> there is the output:
>  1172 have been built (with src.rpm)
>  1662 could be built
>  1118 are in both list
>  544 are not built
>  54 have been built but are not on the list
>
> I am therefore not running the script to build the last 544.
>
> Now what we need is VM/builder. We need 1 Fedora 12 i686 VM running to
> generate the i686 RPMs, and we need 2 CentOS VM one of each
> architecture.
> - Allen, you said you might be able to set up a VM. Is this still ok ?
> Which architecture can you do ?
> - Dan, you said you might be able to run some build on your server
> running CentOS. Is this still ok ? Which architecture does it have ?

Should be OK if you can give me a clear set of steps to follow (sorry!).

Here I have access to:

* x86_64 GNU/Linux with CentOS release 5.2 (Final) Red Hat Enterprise
Linux Server release 5 (Tikanga)

* i686 i386 GNU/Linux with CentOS release 5.4 (Final)


> - Can anyone build the missing VM/builder ?
>
> Once we have this, building the RPMs will take at worst 2 days I think.
>
> This mean that by the middle of next week we can already give access to
> 1662 RPMs.
> So the question remains, where do we set up the repo ?

For reference, here is Jeff's summary of the projects usage requirements:

        "Correct me if I'm wrong, but Pierre estimates that his ~1000
RPMs for an i386 build of the R/rpm repo takes up ~6 GB, and it could
be 3-4 times that, eventually. So, with 18-24 GB per platform (i386,
i686, x86_64, etc.), I'd say 50 GB would be the amount of space we'd
want to begin with for the R/rpm repo."

Right now we need about 20 GB, which seems to be available right now
on bifx.org (correct me if I'm wrong Jeff).


> - Bioconductor.org is for sure willing to support but only up to < 3GB
> until summer.
> - Allen, you said that you might have some resources for that. Still
> valid ? Would it be available on Monday ?
> - Dan, you said that you have plenty of space. Still valid ? Is it
> available on Monday ? :-D

Sorry, I'm not sure where the confusion came from. I can't provide any
hosting personally. Perhaps I was talking about bio.cc? If you want a
shell account on bio.cc I can easily set one up for you. You are free
to use that box (~4 Tb) it in whatever way you like -
http://groups.google.com/group/biocc-server-interface.


> For the rest, it seems to me that we are getting close to have something
> nice up and running.

So far so good!

All the best,
Dan.


> Regards,
>
> Pierre