Re: Internal "mirror" for fou4s

On Thu, Jan 29, 2004 at 10:20:07AM +0100, Markus Gaugusch wrote:
> On Jan 29, Bernd Felsche <[email protected]> wrote:
> > I've finally bitten the bullet and "done it right".
> >
> > The FAQ page mentions a couple of methods of keeping several hosts
> > updated from a common archive.
> >
> > One option not mentioned is to set up a "mirror" using an httpd.

> Your apache setup seems to be ok.

> > On each client, run
> > 	fou4s --export client.fou
> > and transfer the file to the master server.

> Uhhhmm ... no. The import/export feature was thought for completely
> offline hosts. You should rather run "fou4s --host client1", which imports
> the RPM db of that host via SSH.

> > Proceed with updating the master server
> > 	fou4s --auto --checkfou4s

> No, you must run the --auto together with the --host parameter of
> the client. If several clients have nearly the same files, they
> are only downloaded once. I'd probably also don't use --auto but
> fou4s -ued (update patch descriptions, examine, download) and
> redirect to /dev/null. The clients should run fou4s a bit later
> and will just find the new packages, send notification mail, etc.

OK. I'll try that.

	fou4s --host dyn130 --auto -v 
reports an update of the rpm db.

It's downloading patches for the first client now.

> I have yet to think of a mechanism for "publishing" updates, e.g.
> the admin tries an update on his test machine and says "ok, I want
> this on all others".

It's sort of like squid; but with a permanent cache. An fou4s-proxy
server running on a master system and used as the mirror? The proxy
watches requests and either fulfills them from it's cache if
available, or proceeds to download and make a copy of what's
downloaded.

> I must admit, that I don't use such a setup anywhere, because I've
> never had a farm of SuSE machines to handle :(
> Only a couple of machines here and there - still handled manually.

At the momemnt, I'm operating a stud. :-) All up there are eight
SuSE boxes sitting here; 7 of them on 8.2 but with Athlon, P4, P3
and C3 processors coinfigured with different mixes of packages.
Maybe you can figure out why I've invented something different.

The usual population here is four. The other half are being
despatched all over the continent; starting tomorrow. They're all
connected through a WAN, with a single gateway to the rest of the
world.

I didn't want to use NFS because the write access isn't safe;
firewalls shouldn't have write privileges on other systems.
Read-only access to NFS didn't seem right either; WAN bandwidth
won't be stunning as NFS is rather inefficient.

-- 
/"\ Bernd Felsche - Innovative Reckoning, Perth, Western Australia
\ /  ASCII ribbon campaign | I'm a .signature virus!
 X   against HTML mail     | Copy me into your ~/.signature
/ \  and postings          | to help me spread!
--