[Savannah-hackers] Re: What's up?

[Mathieu Roy]

Nic <address@hidden> said:

> Mathieu Roy <address@hidden> writes:
> 
> > Richard Stallman <address@hidden> said:
> > 
> > > I have been occasionally suggesting "Would you like to help run Savannah"
> > > when someone says he would like to help.
> > 
> > Yes, and that's ok.
> > 
> > But please, now at this point, send people to sysadmins.
> > 
> > Currently we have no manpower issue at Savannah. 
> 
> That's not true Mathieu. I have monitored the savannah situation
> since it's inception. I wouldn't say the savannah hackers are
> drowning... but they are struggling.
> 
> 
> > The previous month, I was not able to maintain as I usually do
> > Savannah, that's why we Rudy was a bit exhausted.
> 
> So we had one person who was maintaining. What if he had been ill?
> hit by a truck? bitten by a dog? eaten by aliens?

I think that at the next vacation, Vincent and you will be able to
help. 
This august month was a special case, that cannot be generalized.

 
> > I do not think we need someone else right now. We're at least,
> > including Nic and Vincent, 4 persons, and that's perfectly enough.
> > In the past, we were 3 or 2 and we had lesser useful tools and
> > it's was still working fine.  (I do not mention Loic and Jaime
> > because they have enough work to do with EUCD, courses and are not
> > really active on Savannah - which is not a problem)
> 
> More people helping savannah means less work for all of us and more
> and larger loads. It can't be a bad thing.

Savannah work is not complicated but it takes some times to truly
understand the better way to do things.

It's easier to have a good staff of 3/4 persons that are here for a
whole year than having a beginner staff of 10 persons that are here
for 1 month.
So having too many unexperienced people can be a bad thing.


 
> > What we really need know is pretty clear:
> > 
> >         - A responsive staff of sysadmins to handle the machines that
> >         are not savannah.gnu.org
> >         I think they surely lack of manpower
> >         All the current complains of users are related to that
> 
> I agree with that. But maintaining the GNU machines is a much more
> delicate job than maintaining savannah. The GNU machines have "real"
> walking and talking users that they must serve, as well as projects
> like savannah. Trust is therefore extreemly important.
>
> 
> The work currently being undertaken will improve the situation
> because it will reduce the ad-hoc nature of uploads and reduce
> access to the machines. Less users always means less work.
> 
> But they do need more help.

I think they truly need more help. As I stated before, all the
important pending request from savannah users depends on sysadmins.

It's better to add people while there are obviously lack of people,
even if it's temporary.

> 
> >         - Some hardware. One year ago, I noticed that before the
> >         end of this year we were probably running out of disk
> >         space. To face the situation, I sent a bunch of mails and
> >         I moved some rather unused data to a temporary location.
> >         Since then, I never received any answer about that and
> >         know we're going to run out of disk space soon or later.
> 
> I agree with this as well. In the short term an extra disc or two
> wouldn't go amiss.
> 
> Maybe we can just NFS map some space in?

According to my experience, NFS does not really fit for server with
got many many accesses, we will surely experience new problem not
trivial to fix. 
I think that getting an harddisk would be something easier.

 
> In the longer term we need to move the savannah code to a more
> distributed architecture. Storing sessions on the database and
> having some cvs name -> host server mapping would be a start.
> 
> I am actively tryoing to work out how we could do that. Hopefully the
> RSS project will be a first step.

It would be in interesting step for Savannah.


> > So I think that's manpower at savannah is really not a problem
> > now.
> 
> It's a terrible risk Mathieu, to run a system with this few
> people. If you were taken ill now could Rudy take over? No, he's busy
> with his exams. Could I take over? No, I just don't have that kind of
> time.

I was thinking that you have this time. If you have not, you should
only contribute to the code and the support request.
If people are about to join the savannah hackers, they need to have
some free time to be able to run the server.

It's better, IHMO, to have few admin that we can rely on than 20
persons that can only help from time to time.
I'm not saying that help from time to time is something useless, but
it not enough.

Basically, at this point, I think that the better way to add savannah
hackers is to let people come and join step by step.


> >         I am going to alleiviate this tool problem a little bit by
> >         adding 
> > 
> > You should delay that project until I agree with the CERN people and
> > the Karlsruhe University for the database cleaning. Things will
> > heavily change soon.
> 
> Yep. I got that message. My code is very easy to change.

Good.

 
> > While the savannah code base can be enhanced in many ways, while we
> > should have multiple computers (which is not a problem with the
> > current code base), 
> 
> It is a problem with the current code base. Our scalability plan is
> ever bigger machines. That is obviously a bad scalability plan.

That's not our plan. I proposed several months ago to host the
database on one computer, the PHP interface on one other computer and
the CVS on another one. It would separate stuff that use most of the
CPU (https, anoncvs). 

But since it do not seems to be possible, the other option is to
change the computer.

But personally it would not be a problem to me if we get a new box
only for PHP/database and let the CVS on this one. I think this
computer should able to handle CVS easily.


> > I do not agree at all with the idea that the current system let us
> > to much work to do. It has not worked for two years without proving
> > to be a little bit useful.  There is much to do for the codebase,
> > not for the administration -- and it's not even really a manpower
> > issue.
> 
> It's not a bad system. I'm not critising the savannah code or the
> marvellous work you and all the other savannah hackers have done. But
> there is room for improvement. As a fledgling savannah hacker
> (without a lot of time) I feel the need for more tools to monitor the
> situation. With such tools I would feel confident enough that I could
> manage savannah on my own for a week, even with my limited time.

The RSS stuff is indeed a good addition. And you're right, any
enhancement that can save admin time is a good thing. But what takes
more time is the review of the pending projects, and it requires some
experience, despite the fact that we wrote a savannah.el that reduce
per 10 the time it require to do that job.

Regards,


-- 
Mathieu Roy
 
  Homepage:
    http://yeupou.coleumes.org
  Not a native english speaker: 
    http://stock.coleumes.org/doc.php?i=/misc-files/flawed-english