Bon ben merci beaucoup !

Ok, un certain problème résolu, ça tourne maintenant à http://emergency.gna.org

Sans plus de succès depuis la madeleine

J'ai testé le mode actif, j'ai testé en coupant les parefeu depuis la madeleine comme depuis ici même.

Pour ce qui est de passer par un proxy, j'en sais rien, ça m'étonnerais pas. L'endroit où je suis est réputé pour son réseau ultra-chelou au fonctionnement étrange. Ceci dit, ça colle pas depuis la madeleine qui est sans proxy.

After unpacking 1734kB of additional disk space will be used.
Get:1 http://linuxsoft.cern.ch cern/7.3.4/i386/apt/os lftp 2.4.9-2 [699kB]
Fetched 699kB in 0s (955kB/s)
Committing changes...
Preparing... ########################################### [100%]
1:lftp ########################################### [100%]
Done.
[pcphsft06] /home/mroy # exit
exit
[pcphsft06] /home/mroy > lftp ftp.jexiste.fr -u gnaemergency
Password:
lftp gnaemergency@ftp.jexiste.fr:~> ls
ls: Login failed: 530 Login incorrect.
lftp gnaemergency@ftp.jexiste.fr:~>

chelou !

Non. Je viens de tester, ca marche.

Ok, j'ai valide le compte gnaemergency jusqu'en 2010 et j'ai rattache le nom emergency.gna.org.

Il te suffit de mettre tes fichiers sur le FTP (dans le repertoire www) et de faire pointer le DNS vers web.jexiste.org (CNAME).

Thanks for fast answer... take time though to be sure the procedures are robust vs implementation

This time, you had no way of warning the users / project admins (apart linuxfr). Having a list of project admin available for warning by mail would be fine...

My proposal of http://www.gna.org / gna.org that could be differentiated in case of emergency is a simple / easy to remember way of access for users... If you want to have a http://emergency.gna.org : that will do (just put it in the FAQ which may not be accessible when it's down... so it has to be easily remembered). You may put there planned downtime (which do not happen that often)

I agree that a wiki is a place for breakage, as a ML for spam... there will always be bad guys out there :-( The important thing is that there remains a way to contact the admin / be informed, and this place must be known.

BTW, tuxfamily.org had a similar problem today : power outage at lost-oasis ;-) maybe they have the same cleaning lady as linuxfr :-))

I was not far for my previous request, as it was support #218 : reread the (full) discussion...

From a user point of view, as admin of a project, my requirements are :
- to be informed when it happens, when it's restored
- to get a planned date for restoration of service
- be told whether there's data loss or not (and what I should have done to avoid data loss)

Possible implementations are :
- full redondancy on several sites (well quite expensive), quite compulsory for DNS (at least for the secundary)
- no redundancy : means of information are indirect (currently linuxfr is quite good), if you still could have sent a mail to subscribers of gna project ML it would have been better
- partial redundancy : alternate hosts are available and can provide minimal service (see below)

For example, you could have a http://www.gna.org hosted on another server :
- if www is unavailable, people go directly to gna.org
- if gna.org (projects) is unavailable, people go to http://www.gna.org and see what happened
=> it's interesting as it does not require any DNS change (which can be long to propagate...)

For hosting, you may ask several LUGs, maybe one can provide you with limited hosting :
- a homepage for information (users have to find something interesting there, or they will forget this URL...)
- maybe a wiki (as was done with tuxfamily.org...)
- perhaps a ML for emergency situations (that project admins are advised to subscribe to)

Last : for backups, they should be externalized if possible - either on user side : for cvs that's easy (but not for support/bugs...)
- or by server side : maybe a regular rsync (but once gna grows it becomes more and more heavy...)

Maybe there's descriptions of procedures / implementations of this already on the web ? other than what I've already done... http://wiki.eagle-usb.org/wakka.php?wiki=BesoinsUtilisateursHebergement in other situations

We definitely need an emergency homepage, as suggested by Baud123, to be able to inform user easily in case all system go offline. This page must be on another location that gna machines (otherwise it would probably not help much). An account on this machine should be accessible through ssh to gna admins + virtual domain.
We could set emergency.gna.org -- if we use a gna address, it means that the secondary DNS must be a working thing.