Subject: Problems with anoncvs (was Re: My cvs update just took 11:57 (12 hours!))
To: None <firstname.lastname@example.org>
From: Douglas Wade Needham <email@example.com>
Date: 02/06/2004 19:39:58
Quoting Sean Davis (firstname.lastname@example.org):
> On Fri, Feb 06, 2004 at 07:42:04PM +0100, Christian Biere wrote:
> > I don't understand why people want to use the main server at all.
> > The mirrors are usually much faster (that's even more true for ftp).
> Usually? I've never seen a mirror be faster than anoncvs since they
> upgraded. And the old anoncvs server was able to handle a *LOT* more
> concurrent users than this one with its resource limits. It would get slow,
> but my point is it would *work*. This anoncvs server just drops you if there
> are too many users connected. There needs to be more than one anoncvs, with
> some sort of load balancing scheme setup, if they're going to be dropping
I find myself agreeing enough to come out of stealth mode.
Sheesh...maybe the anoncvs.netbsd.org server is running 1.6ZI (only
halfway joking). A few days ago, I put a 1.6ZI kernel up on my AMD K7
XP/2500+ system and tried to rebuild my packages. After it was was
SOOOO SSSLLLOOOWWW (my guess is that it was going to take 10-15x as
long to build as usual), I backed out my 1.6ZI kernel and tried to do
a cvs update. It was bad enough that I had to double check to make
sure I got my 1.6ZH kernel back in place, or that somehow, I had not
gotten onto my ancient Sun 3.
Since I had to do a rebuild at the OSU Radio Observatory (where I
volunteer), several days ago I decided to update the 1.6.2 source
trees I have there from RC4 to rule out network problems on my end.
With Ohio State on Internet2, I have not seen connectivity problems
from there short of a downed router completely isolating a segment of
campus. Unfortunately, the problem persisted, so I started checking
the mailing lists and suspecting a server outage.
Like several folks have written, until recently, I have had no
problems using anoncvs.netbsd.org to keep my checked out repositories
up to date. Indeed, I have been tracking source through CVS at least
once a week for the past three years. A number of times, even here in
the past few months, I was getting updates for src and xsrc in about
10 minutes every other or third day at about 4AM ET, even using
specific time (midnight ET). However even this has not worked. Most
times, I just fail to get a connection at all, and if I do, my guess
is that the update would take as long if not longer than the 11h57m
mentioned at the beginning of this thread.
And speaking of mirrors. The few times I have used a mirror, I found
it was either slower, or massively out-of-date (sometimes almost a
week). This is not helpful when you are trying to track current or a
pending release. Worse, I only see one mirror in North America now.
This means that if I have problems with it, I must use the
intercontinental links with the higher latency involved in those
links. Of course, if anoncvs is going to be this bad, perhaps a 6
hour update is better than a 12 hour one or none at all.
Douglas Wade Needham - KA8ZRT UN*X Consultant & UW/BSD kernel programmer
Email: cinnion @ ka8zrt . com http://cinnion.ka8zrt.com