On Sat, 2008-01-26 at 15:58 -0500, Derek Tattersall wrote: > macgyver wrote: > > > > When I had two servers running Fedora2 and both my wife and I's > > workstation, and my laptop all running FC2 - what I did was get one of > > the servers to do an automatic update at say 02:00 - and do not delete > > the resultant packages... > > > > The resultant packages were then NFS shared to the rest of the systems > > as a background mount..... > > > > The result - when yum ran on the rest of the machines, if it found those > > packages already on disk - it didn't re-download them - thus saving > > downloading everything 5 times.. > > > > Also no problems with keys - as still using the repo ;-) > > > > Ok - didn't stop the download of what needed to be downloaded - but that > > traffic was minimal compared to the actual packages..... > > > > Helluva lot better (IMHO) that rsyncing the entire repo - 'cause you > > won't need a vast majority of what that repo has.. > > > > > > Just my 2c > > AM > > > > > I like that idea. > > I could make a script on one machine that looks like this: > > #!/bin/bash > yum $* > rsync -vRd /var/cache/yum/*/packages/*rpm /var/www/html/yum > > and then use that to either make a local repository, or just copy it to > the /var/cache/yum directories on my other machines. > > I think that would do what I have in mind. My method on all machines...(including server) Moved /var/cache/yum to /var/cache/yum_old mkdir /var/cache/yum export /data/adminfilesyum as read/write NFS share (/etc/exports and service start NFS - and portmapper of course) mount <server>:/data/adminfiles/yum /var/cache/yum No rsyncing, or copying any data at any time - when *any* machine did a yum, it looked to see if the files were in /var/cache/yum - and if not downloaded it. The first machine would d/l the files first - all others then didn't need to - they just used the locally cached copy. now /var/cache/yum wasn't local to each machine - but on a server that was local to them - server was up all the time - so was not an issue.. yeah - opening NFS up and portmapper is theoretically a security issue, but if all machines are in the same LAN - might not be a huge issue. YMMV if you don't have a dedicated "server" machine - but your method above would be copying that data twice - once to the /var/www/html/yum area, and then once again down to the "secondary" machines - and data from the d/l packages can get really rather large over time, unless you are clearing it out - so might be worth updating one machine first, then running the following on the other two "client" machines rsync -av <firsthost>:/var/cache/yum/ /var/cache/yum/ yeah - you'd be copying the xmls and sqlite data - but that was never a problem for me - all machines were actually using the same data via the gift of that NFS mount... again - only my 2c. AM
