[Mageia-sysadm] Infos about the machines

[Michael Scherer]

Le vendredi 08 octobre 2010 à 01:07 +0200, nicolas vigier a écrit :
> Hello,
> 
> Here is some infos about the machines we have, or we will have.
> 
> Currently we have those machines available :
> - one dedibox server :
>   http://www.online.net/serveur-dedie/offre-dedibox-v3.xhtml
> - two gandi virtual machines, one share for each :
>   http://en.gandi.net/hosting/
> 
> Later, we should be able to install some machines in a datacenter
> (thanks to rapsys). It should be at least 4 powerfull machines. Each
> one should have between 4 and 12 cores, 8 or 12GB of RAM, and a lot of
> disk space.
> 
> Before we are able to setup the machines in the datacenter, we can start
> using the machines we already have. They are currently used for this :
>  - the dedibox is used to setup the forum
>  - the first gandi VM is used by :
>    * meetbot
>    * a temporary svn server (to host the website, and servers config)
>    * an ldap server (Buchan is working on it)
>  - the second gandi VM is not used yet, but should be used to host the
>    main website soon
> 
> I think we can probably start soon to setup the following services :
>  - dns servers (one on a gandi VM, an other one on the dedibox)
>  - mailing list server on ml.mageia.org on the dedibox as we may need
>    new lists soon to organize teams. If using sympa, with users managed by
>    the ldap server, we need the ldap server and its web interface ready.
>    But we can start testing now.
>  - main mageia web site (on the second gandi VM)
>  - epoll
> 
> Later, the machines in the datacenter can be used for this :
>  - Server1: bugzilla/nagios/dns1/sql/ldap/api/mail/mailling/pastebin/wiki/planet
John have working on updating planet on zarb.org, i guess he will be
able to transfer that.

>  - Server2: svn/git/BS scheduler/hdlists/primary mirror
is this were the key to sign packages will be kept ?

>  - Server3: BS node
>  - Server4: BS node

Does people have direct access on both ?

> If you have any other ideas, comments or questions, don't hesitate to
> reply.

No backup server ? No postfix ( primary and secondary ) ? No wiki ?
Would all web applications be hosted on the same server ( ie epoll,
transifex and other applications ? )

We should have a secondary ldap. I would also add a ticket system which
is not bugzilla ( as infrastructure as a product would be weird ).

And we may need somewhere to write the doc, if possible something that
can be used offline.

We also need to discuss what is our responsibility and what is not ( ie,
who is root on what server, mainly the website one, and who decide of
the various setting, mainly php/apache ). I would propose that we
leverage a VCS + some soft like cfengine/puppet to delegate some part
( like some vhosts settings on some server ) to some others groups ( and
this would also provides tracability, ie, no direct root access ).  


-- 
Michael Scherer