just to beat this savagely to death (or "overdiscuss" it :-), before i tried a full "yum update", my /usr filesystem was 6G and was about 5.5G full. so i still had half a gig to play with, which seemed plenty at the time. when i checked "yum list", i was told i had about 350 packages that could be updated. so i ran "yum update", which failed, telling me i was about 250M short of space in /usr. curious, i thought, since i wasn't installing new packages, just updating existing ones and i couldn't believe that newer packages would be that much larger. but, as seth mentioned, yum apparently takes its space diagnostics from rpm, and i don't know how rpm is deciding what is and is not sufficient space. so i started updating a chunk at a time: yum update "a*" yum update "b*" just to see what would happen. turns out this was one solution -- doing updates in smaller pieces worked fine, and "df" afterwards showed little change in the usage of /usr. updating the numerous "kde" packages was a stumbling point, so i just added exclude=kde* to yum.conf for the time being, and finished the update in one operation. end result was that i still had 400M left under /usr (yes, i know that's getting tight, i'll shift things around later.) so, finally, i figured to update all the kde stuff, removed that line from yum.conf, "yum update", only to be told that i was all of 18M short in /usr. argh, so close! ok, so exclude=kde-i18n* and "yum update" is off and running. when it's done, i still have 400M available in /usr, so i remove the exclude statement, run "yum update" and everything works. when that's done, i still have 405M of /usr to spare and a fully-updated system. the moral of the story? that "out of space" warning certainly has the potential to bite people who don't know about it and have a lot of pending updates. rday
