new server migration

[Havoc Pennington]

Hi,

We have the new server machine and lots of people have accounts on
there already. If you haven't sent me your ssh key please do so, as
we'll be cutting CVS over at some point.

To migrate, we need a plan. There's been a fair bit of #freedesktop
discussion but I want to be sure everyone is "in the loop" at least on
a high level.

Here are the current services:

 - the CVS repository
 - the web site
 - the cron job that checks the web site out of CVS and builds
   the final static pages
 - the web site includes a lot of software release tarballs
 - the cron job that updates the D-BUS documentation 
   (this job runs on my workstation currently)
 - the mailing lists (hosted on listman.redhat.com, so separate 
   from the current cvs/web machine)
 - DNS and mail aliases (hp at freedesktop.org, xdg-list at freedesktop.org,
   etc.)
 - backup

People are discussing adding Wiki functionality for content managing
the web site either during or after migration. 
http://pdx.freedesktop.org/twiki/ shows what this would be like 
(though that demo site has the major problem that it breaks 
all links, not acceptable I don't think)

I would like to see an overall plan for how we'll do the migration.  I
might propose that we do one service at a time, though some of this
could be in parallel:

 1. write down a process for making sysadmin changes
    so it isn't mass chaos - admin changes should be logged, 
    posted to a list, or something along those lines. 

 2. get backups running and verified

 3. move the CVS repository, setting up cvs.freedesktop.org pointing 
    to it.

    update cron jobs on current server to pull the web site from the 
    new repository location.

 4. move the current web site, unmodified. Move cron jobs
    to the new server.

 5. set up mailman on new server and import subscriber lists 
    from listman.redhat.com

 6. add Wiki functionality, taking care to preserve existing URLs 
    and content

 7. move to a saner way to do software releases (don't check them 
    in to CVS) - possibly take scripts from gnome.org, as gnome 
    has already addressed the issue of lots of little tarballs

 8. create a bugzilla instance (I'd like to see some of the kde.org
    bugzilla changes, too, the kde bugzilla is much simpler to use)

Havoc