[Savannah-hackers] Re: [gnu.org #218517] webpages sync

[Sylvain Beucler via RT]

On Mon, Jan 03, 2005 at 11:31:49AM -0500, James Blair via RT wrote:
> > address@hidden - Mon Jan 03 11:18:35 2005]:
> > 
> > No, it was occuring for several weeks already :/
> > 
> 
> It occurs to me that we may not have proper error handling for this. 
> www.gnu.org has been down several times recently, and I don't think our
> system takes that into account.  If this is an intermittent problem,
> that may be the cause.

Indeed, in the Savannah reports, I have seen a lot of "checkout
complete", but no error :/


> Maybe the creation of the website should be de-coupled from creating the
> cvs repository.  Instead, maybe a flag of some sort (in the db or on the
> filesystem) is needed indicating whether the website creation succeeded
> or not.  Then the cron job would just create any websites that haven't
> been created yet.

Well, Savane is designed to work by remotely accessing the database
(which means there are several "backend" installations), and thus the
Right Way would be to install the Savane backend at www.gnu.org; it
will then update the websites depending on whether they are activated
in the database. Since the process is ran directly at the target
machine, the error handling can be done precisely.

I think that's the way it is done at Gna! (with home.gna.org separated
from the other parts of the system). I can ask more information on
that point if you want.

-- 
Sylvain