[WBEL-users] WBEL Vs Centos ? :-S

Kirby C. Bohling kbohling@birddog.com
Mon, 6 Dec 2004 23:01:44 -0600


On Mon, Dec 06, 2004 at 10:43:03PM -0500, William Hooper wrote:
> 
> Kirby C. Bohling said:
> [snip]
> >
> > Just out of curiousity what mirror do you use?
> 
> Normally NCSU, with the understanding that they always seem to need poked
> to get anything new (I had to do it at least once before this new push of
> updates).
> 
> > I checked after 48
> > hours after John said, "it should be trickling out to all the mirrors".
> 
> The difference is, at this point I looked at the main mirror, grabbed the
> updates I needed and put them on my local mirror.

Yes, I could have done that.  I figured the fact that there is no
public reference to /pub is a clue that John would appreciate me not
banging away on it.  I'm fairly sure that if everyone did what you
did (espeically in an automated way), it would crush the beau.org T1.

> 
> > Not a single mirror had any of the updates.  That's
> > happened to me several times.  In all of the cases NCSU was having a
> > problem.  When NCSU solved their problem, everyone else had the problem
> > solved too.
> 
> I never said that wasn't the current situation.  I just said that mirror
> admins should be poked to point out that it isn't a good situation.
> 

I felt you implied it fairly strongly.

> > When ncsu.edu deleted all of the updates, all but 2 mirrors deleted
> > all of their updates (the two of them are in my posts to the list).
> 
> My local mirror didn't :-)  I don't trust NCSU enough to put the
> "--delete" option in.

Yeah, I removed my --delete option after that fiasco.

> 
> > Yes, seeing that they are there in no way shape or form implies that
> > there is a scaleable way to get to them.
> [snip]
> > Really, do you really think it'd be a good idea for everyone to
> > point at whiteboxlinux.org because it's the primary site?  We tried that
> > once before, and as I recall John disabled access to /pub
> 
> Pointing up2date at a server is a lot different from going to it and
> downloading one or two RPMs you need to install.  rhn-applet comes to mind
> really quick.  Cron yum jobs come to mind also.

I didn't use up2date, I use the rhn-applet which reads it's
configuration from the same place up2date does
(/etc/sysconfig/rhn/sources).  I always think of it as up2date, as
that's the name of the window it pops up is "Up2date Notification
Tool" or some such.

I still don't see that as scalable for getting the RPM's.  See that
there is a problem, yes.  For actually getting the RPM's no, not if
everyone did it.

> 
> > because the
> > mirrors couldn't get the files.  I also disagree that NCSU isn't the main
> > source for updates.  In my experience, it is the primary site for scalable
> > access.
> 
> You previous statement "the mirrors couldn't get the files" implies that
> at some point NCSU wasn't the "primary site" that other mirrors used.  The
> choice to change that seems to me (from the outside) to have been made by
> the mirror admins.

That's a singular/plural problem on my part.  At the time, John
said, "I'm cutting off public access to /pub so the mirrors can
catch up", if I am remembering the e-mail correctly.  I have no
direct evidence that at any point more then a single site has ever
used mirrored the site.  It would be a mistake to leap from my
plural usage to saying that multiple places used to mirror directly
from the primary site.

> 
> [snip]
> 
> > You might really want to check out who is mirroring from whom.
> > There is pretty ample circumstancial evidence that nearly everyone
> > is mirroring from NCSU.
> 
> And John has stated on the -devel list that he would like to change that. 
> Not much he can do with other peoples servers, though.
> 

Yes and no.  Several thoughts jump to mind.  I'll bet money the
reason that they use ncsu is just a barrier to entry problem.  The
barrier to mirroring NCSU was knowing how rsync works.  The barrier
to getting mirror access from whiteboxlinux.org meant sending an
e-mail and possibly being rejected.  Having to deal with John, and
adding a password, that might have to be maintained, or dealing with
John if the IP address or DNS name ever changed on the host.  I'd
much rather deal with NCSU (assuming I was unaware that they ran
such a defective mirror).

I'm willing to bet that none of the mirror admin's follow the -devel
list.  Maybe someone (me?) should try sending that to the various
mirror admins a request.  If John publically stated how he would
provide them access (via rsync and putting an IP in hosts.allow for
example), I'd contact the mirror admins to see if they would respond
with an IP address.  I don't want to blindly poke at them, without
some concrete way of knowing I wasn't wasting their time.

If I was John for a day, I'd add these IP's to my hosts.allow for
rsync access:

193.120.14.241  # ftp.esat.net
64.240.156.196  # ftp.kspei.com 
192.35.244.50   # (Antonin.Sprinzl@tuwien.ac.at) gd.tuwien.ac.at
193.190.198.20  # (ftpmaint@belnet.be) ftp.belnet.be

These are the publically given IP address for the mirrors that have
rsync access from them (the others might use rync, but they don't
allow me public rsync access).  Then send an e-mail to the last two
explaining the situation, and which IP's had direct rsync access.

Using Whois, for ftp.esat.net, my guess is that one of these two
e-mail addresses will work work as a starting point:
dave@esat.net, noc@esat.net
(Found using: whois -h whois.ripe.net 193.120.14.241)

For ftp.kspei.com, I'd contact these people:
dnsadmin@lanscape.net
(Found Using this: whois -h whois.opensrs.net kspei.com)

That would at least get you pointed in the correct direction.

	Thanks,
		Kirby