On Wed, Mar 26, 2003 at 07:30:04AM -0500, Daniel Jarboe wrote: > So am I really looking at a maintenance image to update a /usr (which > will later be shared RO among guests), and then copying rpm-written > files from that maintenance image's /var and /etc to each of the images? > > Or is there a way I can rpm -Uhv on each of the images though /usr is > shared RO? Or maybe hack up the rpm file to exclude anything in the > /usr tree for the images sharing /usr. Is there a converse of rpm2cpio, > or do RPMS have to be rebuilt from a .spec and source?
It's a whole lot easier to get two scratch disks the size of /usr. Make a copy of the original contents of the shared /usr (U0) on one, which we will call S1. Make an additional copy on the other, S2. Now: reipl your template image using S1 as its /usr. Update it. Now, start updating your guests. For each guest (this can only be parallelized up to the number of different S2s you have), attach S2 read-write. Use mount with the --bind parameter to mount S2 over /usr, and then apply your service. Shut down the guest. Reimage S2 from U0. Change the guest's directory entry to point at S1 rather than U0. ReIPL it. Repeat for each guest. Adam
