Tony Nelson wrote:
At 6:55 PM +0000 10/28/07, Mike C wrote:
Timothy Murphy <tim <at> birdsnest.maths.tcd.ie> writes:
I want to yum-update on several machines,
and I want to avoid repeatedly downloading the same packages.
So what I want when I run "yum update" on machine A
is for it to look for RPMs in a specified directory
What I do is have a small bash script that rsync's the packages and
headers in the /var/cache/yum area from the one machine, where they are
downloaded during its own update, to the others.
Then I run a normal yum update on the other machines in the LAN after the
rsync, and they will get most of the rpm files from the rsync'ed data, but
it leaves the machines free to download any extras from the external
repos. Not all machines have the same set of rpms necessary unless they
are setup in identical fashion.
By doing it this way the files in /var/cache/yum use at least an order of
magnitude less disk space than if every rpm were stored on the main
machine as a fedora repo. As far as I remember the full set is around
10GB, whereas the machines I update generally only need less than 1GB even
if they are updating after a first install.
This sounds like a really workable solution, and more efficient than the cp
version.
Others suggest squid. Squid or any other proxy won't work well unless the
same mirror is used every time.
Yum-presto on all machines can provide further download savings.
Another solution (similar to rsync) is to use nfs-shared cache
directory. In my case I have /var/cache/yum as a link to this shared
directory. Then I need only one set of updates. With this configuration
lock file is also shared, so you probably can't do 'yum update' on more
then one system at a time.
I am using distributed shell (dsh) for yum updates anyway, so it is not
an issue.
