On Fri, 2004-12-31 at 10:22 -0500, Mail Lists wrote: > On Thu, Dec 30, 2004 at 10:36:29PM -0800, Jamie Zawinski wrote: > > - Is there some way to make yum cache all that crap? I'd be > > http://www.redhat.com/mailman/listinfo/fedora-devel-list > > (speed of yum discussion exceised). > > > Related yum question on (lack of?) speed when large number of > updates to do. > > After doing a fresh install, the first yum update can > take (for me) hours - this is after I have pre populated > the local cache (/var/cache/yum) from another just > installed/updated machine - so no downloads > to do at all. The number of updates to be done is of course quite > large at this point. > > By breaking this update up into a few separate upates > > e.g > yum update x*; yum update kern*; yum update g*; yum udpate > > The time taken seems to be very substantially less (human perception > unfortunatly I have no benchmarks). > > I wonder if there is some inefficiency with the dependency tree > when the number of updates is large that is reduced by > hand breaking the updates into smaller chunks. > You can time the dep resolution. yum -d4 update look for two strings like this. 1104511430.1813159 They're the output of time.time() they are the start and end time of the dep resolution process. in very large transactions I've found that most of the time is spent in the 'testing transaction' phase. That's entirely inside rpm at that point though. -sv
