On Thu, 31 Aug 2006, Les Mikesell wrote: > On Thu, 2006-08-31 at 08:13 +0300, Panu Matilainen wrote: > >> Reproducing an installation starts to approach a valid reason :) However >> build and file time stamps are not reliable way of doing this, nothing >> guarantees that packages arrive in a given repository in the order they >> are built: for example the vendor might have a heavier testing programme >> for the kernel than some minor package, causing kernel to arrive in the >> repo much later than some other package despite having an older timestamp. >> >> If you want reproducable installations, use versionlock (plugin >> available in yum-utils) on the packageset you tested and forget about >> timestamps. > > Is there documentation available for the various plugins and how > to use them together? For example, given a tested system, how > would you tell a box in a different location to update/install > to the same packages and versions? You can set the versionlock file to be somewhere remote, eg locklist=http://my.main.server.com/versionlock/distro/$releasever or similar. Then you just control that one file, all yum update/install operations will use the versions specified there no matter what other versions are available. > Also, now that the download-only option has been moved out of yum > itself, how do you tell it to pre-fetch the packages you are going to > need (either for this or a normal 'update'), so as to be able to plan > the timing of the actual package installation/updates in a way not tied > to internet bandwidth or health of remote repositories? One way to do "download only" with current yum itself is to set tsflags=test in yum.conf, that way it'll just perform a transaction test but not actually do anything to the system. Or you can write a five-line plugin to make it stop once download completes. - Panu -
