[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Some thoughts



On Sun, 13 Jul 2003, Mattias Wadenstein wrote:

> One suggestion for being able to have testing images that are just a
> simple repoint away from actual release would be to have a similar setup
> in the debian-cd dir as the main archive:
>
> debian-cd/
>   woody/
>   sarge/
>   stable -> woody
>   testing -> sarge
>
> And then just repoint the stable link at release time.

And in each of these directories:
images/3.0_r1/hppa/
          .../arm/
...
jigdo/3.0_r0/i386/
          ../arm/
...
     /3.0_r1/i386/

That is {images,jigdo}/version_release/{alpha,arm,...}/filenames

In fact, just at it is setup on http://ftp.de.debian.org/debian-cd/ unless
someone has a better idea.

Phil: Does this match the cd-master setup, does cd-master even have a real
directory structure that would be usable for this? Should it?

> I'm taking a look if the initial cdimage -> a few mirrors distribution can
> be done with bittorrent or if that is too inefficient/difficult/whatever.

Ok, that didn't work out that good, it eats lots and lots of cpu. It would
probably be nice to have this available for our users and I can set this
up, but during the first critical week of distributing images to servers
it eats way too much cpu compared to delivered bandwidth.

So I'm looking for a few mirrors that can take the initial data from the
new cdimage.d.o in the first few days when there is no public access from
that site. The suggestion for releases is to start with a limited rsync
access for a few (5-10 or so) mirrors for a couple of days to a week to
give these a decent chance of getting the data from the master site, then
opening up for public access.

It would be good if these were well-distributed across the (topological)
world, but the important thing is that they are around and willing to put
in a little work at release time to get the images in and available. Well,
and having lots and lots of bandwidth at a fast computer of course.

/Mattias Wadenstein



Reply to: