On Thu, 11 Nov 2010, Scott Robbins wrote: > On Thu, Nov 11, 2010 at 02:57:58PM +0000, Timothy Murphy wrote: >> John R. Dennison wrote: >>> I think it's a fair bet to say it will be at least 4-6 weeks >>> out. >> When it comes, will I be able to upgrade by "sudo yum update"? Probably a casual one off user will not want to follow this path as it is not deterministic as to the method that works AND migrates all prior applications 'in situ' There are something north of 80 simultaneous packages for very minimal install, in the resulting transaction set from a CentOS 5 updated to current, to a locally rebuilt 6 testing candidate [from the second beta]. NOT stripping out unused leaf nodes increases the liklihood of a unsolveable dependency (more on this in a moment) even more That is, I have attempted working forward from: 0. build a local binary archive and run a createrepo 1. a clean starting point backup 2. stripping the box down to minimal leaf nodes 3. plenty of local drive to store the transaction set packages in 4. let yum solve the transaction set [rpm5 can do this as well, but is out of scope here] 5. download all of them 6. create a MANIFEST with ls 7. take another backup to work forward from on later trials 8. try a: rpm -F `cat MANIFEST` This fails -- perhaps due to a sequencing issue; needed scripts are dying and SElinux is in play, perhaps a non-present dependency. ... does not matter 9. fall back to the step 7 backup 10. try a: rpm -U `cat MANIFEST` This fails as well, but differently Disabling SElinux did not help; in trying a third time with SElinux disabled, this fails with scripts issues Also one loses the conversion to ext4 I can take a hint ;) I concluded that a media based upgrade (which is essentially --nodeps) and in a environment where SElinux can work, and scripts frummage may not matter [... WOULD NOT matter if script actions were idempotent, which they clearly are not] was the way to go, and not via: yum ;) I have other fish to fry until CentOS specific media to QA becomes available ;) A hobby user doing this for recreation, or a HUGE (hundreds of chassis or instances) may be able to amortize the cost of finding a solution, but to me, it is not yet an interesting problem to spend time in advance of QA'ing 'real' CentOS content -- Russ herrold