------- Comment From [email protected] 2017-01-18 12:02 EDT------- (In reply to comment #23) > > (In reply to comment #21) > >> Installed lpar onto 154d dasd drive, using LVM automatic partitioning > >> recipe. > >> After installation and reipl, I did the following: > > > >> $ sudo update-initramfs -u > > > > I understand that things work with this explicit call. > > How does it work with e.g. dracut? One must regenerate the initramfs > to activate the additional drive on boot, no?
yes. > Does one required to call zipl (which calls into dracut to regenerate > initramfs....?!) or does one call dracut? user sequence (in that order): dracut && zipl. "update-initramfs -u" includes the final zipl step which is nice. > > However, users can easily miss this since it's not entirely obvious. > Regular ubuntu/debian users do know to call update-initramfs -u when > fiddling/changing rootfs devices (not the most trivial task to be > honest). > Same as fedora users know to call dracut. IMHO. I guess, I've seen too many cases where the user missed this (or a related) step and therefore I was hoping for an automatic solution. > Ubuntu uses chzdev, which generates udev rules, and our initramfs > hooks copy all of the udev rules into the initramfs as udev is running > in Ubuntu initramfs. OK, I learned that Ubuntu does not do root-fs dependency tracking of z-specific device activation udev rules it includes into initramfs. IIRC, zdev might have such tracking if it was used with dracut instead of update-initramfs. Maybe that's why I was on a misleading track. > The boot continues as soon as rootfs is detected to be available. If > other devices are activated in parallel, it should be mostly harmless. > And the boot will not wait to activate all the things that udev rules > specify in the initramfs. > Thus it's kind of a zero sum game, eventually all chzdev->udev rules > specified devices will be activated, and it does not matter much if > some of them are activated from initramfs or post-initramfs. I've seen it matter, where large installations choked within initramfs because of processing udev rules (and in turn things such as multipath events) for all the other (unnecessary) devices. But that would be a different bug anyway. > I added the hook in place such that chzdev can call it when it needs to. > > But I'm arguing that the hooks that chzdev calls will never be > sufficient when one decides to move their rootfs, on any Linux. > Because at initial "chzdev -e" call time the device that is being > activated is not yet part of the rootfs stack, but will become after > one e.g. expands lvm / mdadm / btrfs / etc on to it. Does that make > sense? It does indeed make sense to me. > > If the bug is now declared invalid as in "user error", then I'm puzzled > > about the reasons for a code change. > > a code change was mostly a red-herring OK, got it. Thanks for making this clear. So the bug is invalid unless someone finds a solution for the problem that root-fs dependencies (can) change after any current chzdev hook. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1641078 Title: System cannot be booted up when root filesystem is on an LVM on two disks To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu-z-systems/+bug/1641078/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
