Bug#644389: suspect non-working mdadm.conf created during install
also sprach pille pille+debian+bug-rep...@struction.de [2011.10.05.2133 +0200]: i completely understand what happens as i'm able to boot the system by manually assembling the array and fixing the problem, but how the wrong hostname slips into the initramfs should be investigated by debians devs. It's not the wrong hostname, it is what was current when the array was created. At least we should make sure this is the case. Could you investigate what the 'hostname' command outputs on the d-i shell, just before creating an array in partman during the install? Does it output hetzner, or does it print the one that you set previously during the same install? Have a look at the mdadm manpage, homehost is the name of this feature. There are three ways forward, really: 1. figure out a way by which the initramfs is updated when the hostname is changed; 2. make sure the admin knows that changing the hostname requires telling mdadm; 3. provide a debconf setting to disable the homehost feature (setting it to ignore). -- .''`. martin f. krafft madduck@d.o Related projects: : :' : proud Debian developer http://debiansystem.info `. `'` http://people.debian.org/~madduckhttp://vcs-pkg.org `- Debian - when you have better things to do than fixing systems digital_signature_gpg.asc Description: Digital signature (see http://martin-krafft.net/gpg/sig-policy/999bbcc4/current)
Bug#644389: suspect non-working mdadm.conf created during install
It's not the wrong hostname, it is what was current when the array was created. yes, but i think it's the hostname when the initramfs was created for the installer. At least we should make sure this is the case. Could you investigate what the 'hostname' command outputs on the d-i shell, just before creating an array in partman during the install? i checked and it behaves like expected: it reports '(none)' up to the point where i set it to 'file'. afterwards it reports 'file'. i installed _without_ RAID and used guided partitioning this time. after the system is installed i took a look into /target and there was the hetzner-mdadm.conf more evidence for the d-i deploys standard initramfs with broken/suspect/unupdated mdadm.conf: 1) initramfs has an old timestamp (2011-08-27) and contains a comment that it was autogenerated that day sha1sum is: a8f016e70843f108ae05ab0ef874eaad231005c8 initrd.img-2.6.32-5-amd64 2) last time i configured two arrays, but mdadm.conf contains three. my problem is, that i cannot find it on the install medium (or a file big enough to hold it, tat actually contains it). There are three ways forward, really: 1. figure out a way by which the initramfs is updated when the hostname is changed; there's only one way and it's updating the initramfs when nessessary (RAID config) and remove it when not needed. 2. make sure the admin knows that changing the hostname requires telling mdadm; 3. provide a debconf setting to disable the homehost feature (setting it toignore). the UUIDs are different, that's why the array don't get assembled. since you should now be able to reproduce (even without RAID) i'll trash those VM images. cheers pille DEVICE partitions HOMEHOST system ARRAY /dev/md/0 metadata=1.2 UUID=f6de5584:d9dbce39:090f16ff:f795e54c name=hetzner:0 ARRAY /dev/md/1 metadata=1.2 UUID=0e065fee:15dea43e:f4ed7183:70d519bd name=hetzner:1 ARRAY /dev/md/2 metadata=1.2 UUID=ce4dd5a8:d8c2fdf4:4612713e:06047473 name=hetzner:2 # mdadm.conf # # Please refer to mdadm.conf(5) for information about this file. # # by default, scan all partitions (/proc/partitions) for MD superblocks. # alternatively, specify devices to scan, using wildcards if desired. DEVICE partitions # auto-create devices with Debian standard permissions CREATE owner=root group=disk mode=0660 auto=yes # automatically tag new arrays as belonging to the local system HOMEHOST system # instruct the monitoring daemon where to send mail alerts MAILADDR root # definitions of existing MD arrays ARRAY /dev/md/0 metadata=1.2 UUID=f6de5584:d9dbce39:090f16ff:f795e54c name=hetzner:0 ARRAY /dev/md/1 metadata=1.2 UUID=0e065fee:15dea43e:f4ed7183:70d519bd name=hetzner:1 ARRAY /dev/md/2 metadata=1.2 UUID=ce4dd5a8:d8c2fdf4:4612713e:06047473 name=hetzner:2 # This file was auto-generated on Sat, 27 Aug 2011 19:16:33 + # by mkconf 3.1.4-1+8efb9d1
Bug#644389: suspect non-working mdadm.conf created during install
Package: mdadm Version: 3.1.4-1+8efb9d1 Severity: critical Tags: d-i Justification: breaks the whole system during install of squeeze i partitioned my disks and setup different MD-RAID devices using the partition tool. the corresponding /dev/mdX devices were created properly and i went on creating LVM ontop of it and successfully installed the system. first boot into my new system failed because the root FS could not be found, leaving me in a initramfs-shell. i assembled my RAID devices by hand and compared /etc/mdadm/mdadm.conf (inside initramfs) with 'mdadm --examine --scan', which listed completely different UUIDs. the count of my MD-devices was ok, but all names were prefixed with hostname 'hetzner'. after booting, fixing (the systems) /etc/mdadm/mdadm.conf and update-initramfs all went fine. i don't remember the exact content, but install generated a mdadm.conf like: ARRAY /dev/md/0 metadata=1.2 UUID=12345678:12345678:12345678:12345678 name=hetzner:0 ARRAY /dev/md/1 metadata=1.2 UUID=24680135:24680135:24680135:24680135 name=hetzner:1 ARRAY /dev/md/2 metadata=1.2 UUID=abcdef12:abcdef12:abcdef12:abcdef12 name=hetzner:2 (UUIDs are made up) disks used were new empty and had no affiliation to hetzner online AG you can search the net for 'name=hetzner: mdadm.conf' to yield probably the original UUIDs. pille --- initrd.img-2.6.32-5-amd64: 47498 blocks 3718babf31303bc28114430ec0182e14 ./etc/mdadm/mdadm.conf d24ed7389311fb002ff017c8acb3a032 ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/dm-mirror.ko 4818680648975a4d4181181df95de446 ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/dm-log.ko bb013669544366b71ebd8b335a7d8b0a ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/dm-mod.ko ee2079d7e0bf4da6b64af9e579f6c0ae ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/raid456.ko b210dd35ca38b2e33637351a979561c9 ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/multipath.ko 0c8325eb1abcf4ad33e45a20d44c5bf8 ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/dm-crypt.ko f25a213d65a481b66cfdbf8f6453aacf ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/dm-region-hash.ko cc5177855661025115a2ef74d0d394d6 ./lib/modules/2.6.32-5-amd64/kernel/drivers/md/raid0.ko -- System Information: Debian Release: 6.0.2 APT prefers stable APT policy: (990, 'stable'), (500, 'stable-updates'), (500, 'testing') Architecture: amd64 (x86_64) Kernel: Linux 2.6.32-5-amd64 (SMP w/2 CPU cores) Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Versions of packages mdadm depends on: ii debconf 1.5.36.1 Debian configuration management sy ii libc6 2.11.2-10Embedded GNU C Library: Shared lib ii lsb-base3.2-23.2squeeze1 Linux Standard Base 3.2 init scrip ii udev164-3/dev/ and hotplug management daemo Versions of packages mdadm recommends: ii exim4-daemon-light [mail 4.72-6+squeeze2 lightweight Exim MTA (v4) daemon ii module-init-tools3.12-1 tools for managing Linux kernel mo mdadm suggests no packages. -- debconf information: mdadm/autostart: true mdadm/mail_to: root mdadm/initrdstart_msg_errmd: mdadm/initrdstart: all mdadm/initrdstart_msg_errconf: mdadm/initrdstart_notinconf: false mdadm/initrdstart_msg_errexist: mdadm/initrdstart_msg_intro: mdadm/autocheck: true mdadm/initrdstart_msg_errblock: mdadm/start_daemon: true -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#644389: suspect non-working mdadm.conf created during install
tags 644389 moreinfo unreproducible thanks also sprach pille pille+debian+bug-rep...@struction.de [2011.10.05.1458 +0200]: i assembled my RAID devices by hand How? but all names were prefixed with hostname 'hetzner'. What names? You mean the name= attribute in the mdadm --examine output? disks used were new empty and had no affiliation to hetzner online AG you can search the net for 'name=hetzner: mdadm.conf' to yield probably the original UUIDs. Please provide the exact original and new UUIDs, don't expect me to search the Net for them (I am offline right now). Also, did you change the hostname? -- .''`. martin f. krafft madduck@d.o Related projects: : :' : proud Debian developer http://debiansystem.info `. `'` http://people.debian.org/~madduckhttp://vcs-pkg.org `- Debian - when you have better things to do than fixing systems digital_signature_gpg.asc Description: Digital signature (see http://martin-krafft.net/gpg/sig-policy/999bbcc4/current)
Bug#644389: suspect non-working mdadm.conf created during install
i assembled my RAID devices by hand How? something like: mdadm --assemble /dev/md1 /dev/sd[abcd]4 but all names were prefixed with hostname 'hetzner'. What names? You mean the name= attribute in the mdadm --examine output? excactly. Please provide the exact original and new UUIDs, don't expect me to search the Net for them (I am offline right now). unfortunately i cannot provide you with those. if i could i would have done so, already. the install was a week ago and i don't have any backups of a broken config. don't expect me to memorize UUIDs. Also, did you change the hostname? no. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#644389: suspect non-working mdadm.conf created during install
also sprach pille pille+debian+bug-rep...@struction.de [2011.10.05.1819 +0200]: Please provide the exact original and new UUIDs, don't expect me to search the Net for them (I am offline right now). unfortunately i cannot provide you with those. if i could i would have done so, already. the install was a week ago and i don't have any backups of a broken config. don't expect me to memorize UUIDs. Also, did you change the hostname? no. In this case there is very little I can do. I have done multiple squeeze installs with RAID, and upgrades, and I have not seen this problem. Unless you can reproduce this bug, I cannot do anything but keep this report open. -- .''`. martin f. krafft madduck@d.o Related projects: : :' : proud Debian developer http://debiansystem.info `. `'` http://people.debian.org/~madduckhttp://vcs-pkg.org `- Debian - when you have better things to do than fixing systems the mind of the thoroughly well-informed man is a dreadful thing. it is like a bric-à-brac shop, all monsters and dust, with everything priced above its proper value. -- oscar wilde digital_signature_gpg.asc Description: Digital signature (see http://martin-krafft.net/gpg/sig-policy/999bbcc4/current)
Bug#644389: suspect non-working mdadm.conf created during install
In this case there is very little I can do. I have done multiple squeeze installs with RAID, and upgrades, and I have not seen this problem. Unless you can reproduce this bug, I cannot do anything but keep this report open. i've just reproduced such a failing install inside a KVM. see (combined) screenshot attached from first boot. i've snapshotted the VM at different stages and can provide you with one that just boots for the first time (snapshot3, see screenshot). tell me if you're interrested in investigating it. here's a log of what i did: booted http://cdimage.debian.org/debian-cd/current-live/amd64/iso-hybrid/debian-live-6.0.2-amd64-rescue.iso (sha1=0e5d9409719e657de4cbd9f355545cdc4c869038) expert text install selected some udebs (parted, reiserfs, squash, virtio) hostname = file.lan partition manual for all 4 disks: GPT p1: 6MB BIOS BOOT p2: 250MB RAID1 BOOT p3: 768MB SWAP p4: 3.3GB RAID6 VG/PV configure sw raid: md0: RAID1 /dev/sd[a-d]2: /boot (ext2) md1: RAID6 /dev/sd[a-d]4: LVM PV (VG=vg-sys) configure LVM: LV system: /dev/md1 (4GB) / (reiserfs) snapshot1 install system (normal) install grub finish snapshot2 reboot snapshot3 cheers pille attachment: screenshot.png
Bug#644389: suspect non-working mdadm.conf created during install
On 05.10.2011 22:53, pille wrote: In this case there is very little I can do. I have done multiple squeeze installs with RAID, and upgrades, and I have not seen this problem. Unless you can reproduce this bug, I cannot do anything but keep this report open. i've just reproduced such a failing install inside a KVM. see (combined) screenshot attached from first boot. i've snapshotted the VM at different stages and can provide you with one that just boots for the first time (snapshot3, see screenshot). tell me if you're interrested in investigating it. here's a log of what i did: booted http://cdimage.debian.org/debian-cd/current-live/amd64/iso-hybrid/debian-live-6.0.2-amd64-rescue.iso (sha1=0e5d9409719e657de4cbd9f355545cdc4c869038) expert text install selected some udebs (parted, reiserfs, squash, virtio) hostname = file.lan partition manual for all 4 disks: GPT p1: 6MB BIOS BOOT p2: 250MB RAID1 BOOT p3: 768MB SWAP p4: 3.3GB RAID6 VG/PV configure sw raid: md0: RAID1 /dev/sd[a-d]2: /boot (ext2) md1: RAID6 /dev/sd[a-d]4: LVM PV (VG=vg-sys) configure LVM: LV system: /dev/md1 (4GB) / (reiserfs) snapshot1 install system (normal) install grub finish snapshot2 reboot snapshot3 So it looks like during install time, in d-i, the system were named hetzer (most likely due to DHCP), and all md arrays were created using that name. And later on, it fails to boot with actual hostname. Note that mdadm.conf in initrd contains name hetzer, while actual arrays are named file. Is your real hostname file ? /mjt -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#644389: suspect non-working mdadm.conf created during install
So it looks like during install time, in d-i, the system were named hetzer (most likely due to DHCP), and all md arrays were created using that name. And later on, it fails to boot with actual hostname. i' don't know if it's the initramfs shipped with d-i, but today i did an offline install. no net, no DHCP, no hetzner datacenter. i completely understand what happens as i'm able to boot the system by manually assembling the array and fixing the problem, but how the wrong hostname slips into the initramfs should be investigated by debians devs. Note that mdadm.conf in initrd contains name hetzer, while actual arrays are named file. that's what i'm talking about. Is your real hostname file ? yes. pille -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org