Public bug reported: I've been testing booting an Ubuntu 15.04 disk image written onto an Intel RSTe RAID drive configured for RAID 1. My systemd package version is 219-7ubuntu6.
This type of RAID is known as a "fake raid" or external metadata RAID, configured through mdadm. In order to get this booted at all, I had to do the following: e2label /dev/md126p1 / apt-get -y install mdadm // need to install mdadm of course rm /etc/mdadm/mdadm.conf vi /etc/default/grub.d/dmraid2mdadm.cfg // blank this file and save vi /etc/default/grub // add the lines GRUB_DEVICE_LABEL=/ and GRUB_RECORDFAIL_TIMEOUT=0 and save update-grub vi /etc/fstab // name filesystem LABEL=/ and save I added the GRUB_RECORDFAIL_TIMEOUT=0 option because the first boot always fails. It's unable to run fsck against /dev/md126p1. I checked fsck from a fresh install before rebooting, and it shows the partition as clean. After the failure, however, the system automatically reboots with the RAID correctly assembled with the first partition as the root fileystem. On shutdown, however ("shutdown now") systemd reaches the shutdown target, but the task gets blocked and the system never shuts down until I cycle the power. I've attached the messages printed when I shut down and can provide more information as needed. I believe that this happens because mdmon (the md monitoring service, which is necessary for external metadata RAID arrays to function at all) is shut down prior to something, perhaps a logging service, that attempts to write to the disk. Some side notes: I have not tested the shutdown process thoroughly with 14.04 yet, but I believe it works just fine there. Getting booting from fake RAID to work at all is very tricky on 12.04 and involved changes to the initramfs. I plan to file a bug and perhaps a patch against 12.04 for that. Thank you for your time. Update: after rebooting back into another OS from which I can access the partition Ubuntu's installed on, fsck does have to do a minor repair to the journal. This does not appear when running on a fresh install before booting and shutting down: bash-4.2# fsck /dev/md126p1 fsck from util-linux 2.26.1 e2fsck 1.42.10 (18-May-2014) /: recovering journal Setting free inodes count to 467117 (was 467034) Setting free blocks count to 1748100 (was 1748085) /: clean, 57171/524288 files, 348540/2096640 blocks ** Affects: systemd (Ubuntu) Importance: Undecided Status: New ** Tags: systemd-boot ** Attachment added: "1504_bad_shutdown.txt" https://bugs.launchpad.net/bugs/1490010/+attachment/4453928/+files/1504_bad_shutdown.txt ** Description changed: I've been testing booting an Ubuntu 15.04 disk image written onto an Intel RSTe RAID drive configured for RAID 1. My systemd package version is 219-7ubuntu6. This type of RAID is known as a "fake raid" or external metadata RAID, configured through mdadm. In order to get this booted at all, I had to do the following: e2label /dev/md126p1 / apt-get -y install mdadm // need to install mdadm of course rm /etc/mdadm/mdadm.conf vi /etc/default/grub.d/dmraid2mdadm.cfg // blank this file and save vi /etc/default/grub // add the lines GRUB_DEVICE_LABEL=/ and GRUB_RECORDFAIL_TIMEOUT=0 and save update-grub vi /etc/fstab // name filesystem LABEL=/ and save I added the GRUB_RECORDFAIL_TIMEOUT=0 option because the first boot always fails. It's unable to run fsck against /dev/md126p1. I checked fsck from a fresh install before rebooting, and it shows the partition as clean. After the failure, however, the system automatically reboots with the RAID correctly assembled with the first partition as the root fileystem. On shutdown, however ("shutdown now") systemd reaches the shutdown target, but the task gets blocked and the system never shuts down until I cycle the power. I've attached the messages printed when I shut down and can provide more information as needed. I believe that this happens because mdmon (the md monitoring service, which is necessary for external metadata RAID arrays to function at all) is shut down prior to something, perhaps a logging service, that attempts to write to the disk. Some side notes: I have not tested the shutdown process thoroughly with 14.04 yet, but I believe it works just fine there. Getting booting from fake RAID to work at all is very tricky on 12.04 and involved changes to the initramfs. I plan to file a bug and perhaps a patch against 12.04 for that. Thank you for your time. + + Update: after rebooting back into another OS from which I can access the + partition Ubuntu's installed on, fsck does have to do a minor repair to + the journal. This does not appear when running on a fresh install before + booting and shutting down: + + bash-4.2# fsck /dev/md126p1 + fsck from util-linux 2.26.1 + e2fsck 1.42.10 (18-May-2014) + /: recovering journal + Setting free inodes count to 467117 (was 467034) + Setting free blocks count to 1748100 (was 1748085) + /: clean, 57171/524288 files, 348540/2096640 blocks -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1490010 Title: first boot fails and shutdown fails on vivid with fakeraid To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1490010/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs