One can (to some extent) determine how up to dated a server is by
importing urllib2 and adding the following to the try block in
selectOptimalMirror(mirrors) from the yummy code that was posted a few
days ago.
fo = urllib2.urlopen(url)
hdr = fo.info()
fo.close()
modTime = hdr["Last-Modified"]
print modTime
I notice the Last-Modified time for "mirror"/core/development/headers/
header.info tended to cluster around three general points.
Wed, 25 Feb 2004 11:50:51 GMT
Thu, 19 Feb 2004 11:44:44 GMT
Tue, 24 Feb 2004 11:20:34 GMT
There were also a bunch of other points that most likely reflected an
incorrectly --time flag in rsync.
This implies that the server are mirroring three distinct development
states. This affect yum failover in an intresting manner. One may not
be getting the most recent packages. Consider
baseurl=http://mirror.hiwaay.net/redhat/fedora/linux/core/development//
headers/header.info #date Wed, 25 Feb 2004 11:50:51 GMT
http://distro.ibiblio.org/pub/linux/distributions/fedora/linux/core/
development//headers/header.info #date Tue, 24 Feb 2004 11:20:34 GMT
http://redhat.linux.ee/pub/fedora/linux/core/development//headers/
header.info #date Thu, 19 Feb 2004 11:44:44 GMT
Using the above baseurl and failovers, the first two mirrors may timeout
and yum ends up grabbing header.info from the third mirror. For the
rest of the iteration, yum will be searching for the out of date
packages listed there. It will ignore the more recent packages on
mirror one and two. hmmm.
I started looking into this in order to prevent urlgrabber from
downloading header.info every session. It looked straight forward to do
a comparison between between the local .info time and the remote .info
time. I also thought Tomas Junnonen could use this to reflect how up to
date a mirro is.
It presents a possible project for a happyYumUser who doesn't write
code. Contact the sysAdmins for mirrors that have the --time flag
incorrectly set for rsync.
Dave Farning
