>>> On 8/29/2012 at 03:18 AM, "Pavelka, Tomas" <tomas.pave...@ca.com> wrote: > My understanding is that zipl does not understand the filesystem the kernel > and initrd images are on (like GRUB does), but instead stores the offset at > which the kernel image can be found (like LILO does).
That is correct. > What happens when the > initrd image ends up on one physical device and the kernel image on another? You get a failed IPL. > Or part on one device and part on another? Will zipl still be able to boot? No. > My guess is that it won't which makes booting from multi disk LVM possible in > some cases, but generally unsafe (which is why I suspect the error message in > the helper script is there). Is this a correct interpretation? It would be > very useful in my project if I was proven wrong here. I would say that things can be guaranteed to work if /boot is a separate file system, is composed of one single LVM "segment" (i.e. contiguous physical blocks on a single PV), and is never resized to be larger. I.e., a setup that mostly negates all the advantages that lead most people to use LVM in the first place. I think that you got lucky with your testing, in that all the blocks that needed to be on one physical device were actually there. The helper script was telling you that it couldn't guarantee that, so it refused to run. Mark Post ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
