raid degraded after update to Karmic beta
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
dmraid (Ubuntu) |
Invalid
|
Medium
|
Unassigned |
Bug Description
After upgrade to Karmic Beta, udev no longer finds /dev/sda1, /dev/sda2, /dev/sdb1 or /dev/sdb2, only devices /dev/sda and /dev/sdb are visible. This also causes one raid (/dev/md0) to disappear and /dev/md1 to be degraded.
However, since system files are on /dev/md1, the system works. Only the RAID array is degraded, so it does not work. The /dev/md1 should be assembled of /dev/sda2 and /dev/sdb2.
Problem seems to be with udev, since kernel sees the missing partitions on boot (according to dmesg):
[ 1.890383] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[ 1.895670] sda:
[ 1.903143] sda1 sda2
...
[ 1.896009] sd 1:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[ 1.896106] sdb:
[ 1.896285] sdc: sdb1 sdb2
Those partitions are also visible with fdisk -l (output attached)
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #1 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #2 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #3 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #4 |
affects: | ubuntu-on-ec2 → udev |
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #5 |
Version of udev is 147~-5
robegue (r087r70) wrote : | #6 |
I'm also encountering this bug after upgrading to karmic (2.6.31-13).
I can still boot using an older kernel (2.6.27-14).
the /proc/mdstat says that the arrays are inactive and 'mdadm --A -s' in the busybox doesn't work. If i stop the md0 with 'mdadm -S /dev/md0' and then reassemble, it is assembled with a single drive instead of two.
In summary the array is marked as degraded even if it's not.
Scott James Remnant (Canonical) (canonical-scott) wrote : | #7 |
Could you run "apport-collect 449876" and also provide the output of "blkid" Thanks
Changed in udev (Ubuntu): | |
status: | New → Incomplete |
importance: | Undecided → Medium |
affects: | udev → null |
robegue (r087r70) wrote : apport-collect data | #8 |
Architecture: amd64
CustomUdevRuleF
DistroRelease: Ubuntu 9.10
Lsusb:
Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 001 Device 004: ID 046d:c016 Logitech, Inc. M-UV69a/HP M-UV96 Optical Wheel Mouse
Bus 001 Device 003: ID 413c:2005 Dell Computer Corp. RT7D50 Keyboard
Bus 001 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
MachineType: System manufacturer System Product Name
NonfreeKernelMo
Package: udev 147~-5
PackageArchitec
ProcCmdLine: root=/dev/md0 ro
ProcEnviron:
SHELL=/bin/bash
LC_NUMERIC=C
PATH=(custom, user)
LANG=it_IT.UTF-8
LANGUAGE=
ProcVersionSign
Uname: Linux 2.6.27-14-generic x86_64
UserGroups: adm admin audio avahi avahi-autoipd backup bin cdrom clamav crontab dhcp dialout dictd dip disk fax floppy fuse haldaemon klog lp lpadmin mail messagebus netdev news nvram plugdev powerdev sambashare scanner src ssh sudo sys syslog tape tty users utmp uucp vboxusers video voice www-data
dmi.bios.date: 09/25/2006
dmi.bios.vendor: Phoenix Technologies, LTD
dmi.bios.version: ASUS M2N-E ACPI BIOS Revision 0402
dmi.board.name: M2N-E
dmi.board.vendor: ASUSTeK Computer INC.
dmi.board.version: 1.XX
dmi.chassis.
dmi.chassis.type: 3
dmi.chassis.vendor: Chassis Manufacture
dmi.chassis.
dmi.modalias: dmi:bvnPhoenixT
dmi.product.name: System Product Name
dmi.product.
dmi.sys.vendor: System manufacturer
robegue (r087r70) wrote : BootDmesg.txt | #9 |
robegue (r087r70) wrote : CurrentDmesg.txt | #10 |
robegue (r087r70) wrote : Dependencies.txt | #11 |
robegue (r087r70) wrote : Lspci.txt | #12 |
robegue (r087r70) wrote : ProcCpuinfo.txt | #13 |
robegue (r087r70) wrote : ProcInterrupts.txt | #14 |
robegue (r087r70) wrote : ProcModules.txt | #15 |
robegue (r087r70) wrote : UdevDb.txt | #16 |
robegue (r087r70) wrote : UdevLog.gz | #17 |
robegue (r087r70) wrote : XsessionErrors.txt | #18 |
Changed in udev (Ubuntu): | |
status: | Incomplete → New |
tags: | added: apport-collected |
robegue (r087r70) wrote : Re: udev causes raid to degrade after update to Karmic beta | #19 |
blkid output:
/dev/sda1: UUID="7ce8485c-
/dev/sda2: UUID="4f0243d1-
/dev/sda5: UUID="a3d5f583-
/dev/sda6: UUID="c478e150-
/dev/sdb1: UUID="7ce8485c-
/dev/sdb2: UUID="4f0243d1-
/dev/sdb5: UUID="a3d5f583-
/dev/sdb6: UUID="c478e150-
/dev/md0: UUID="3e651827-
/dev/md1: UUID="7e5223ba-
/dev/md2: UUID="5ecad822-
robegue (r087r70) wrote : | #20 |
#cat /etc/fstab
proc /proc proc defaults 0 0
usbfs /proc/bus/usb usbfs defaults,
/dev/md0 / ext3 noatime,
/dev/md1 /home ext3 noatime 0 2
/dev/md2 /scratch xfs noatime 0 2
/dev/mapper/md4 /crypted ext3 noatime,
/dev/cdrom /media/cdrom udf,iso9660 user,noauto 0 0
Scott James Remnant (Canonical) (canonical-scott) wrote : | #21 |
From your collected data:
UDEV [1255421703.174825] add /devices/
UDEV_LOG=3
ACTION=add
DEVPATH=
SUBSYSTEM=block
DEVTYPE=partition
SEQNUM=2041
ID_TYPE=disk
ID_BUS=ata
ID_MODEL=
ID_MODEL_
ID_REVISION=3.AAE
ID_SERIAL=
ID_SERIAL_
ID_SCSI_
ID_PATH=
ID_FS_VERSION=
ID_FS_TYPE=
ID_FS_USAGE=raid
ID_FS_UUID=
ID_FS_UUID_
DKD_MEDIA_
MD_LEVEL=raid0
MD_DEVICES=2
MD_UUID=
MD_UPDATE_
MD_EVENTS=1
DKD_PRESENTATIO
DEVNAME=/dev/sda1
MAJOR=8
MINOR=1
DEVLINKS=
UDEV [1255421703.178081] add /devices/
UDEV_LOG=3
ACTION=add
DEVPATH=
SUBSYSTEM=block
DEVTYPE=partition
SEQNUM=2042
ID_TYPE=disk
ID_BUS=ata
ID_MODEL=
ID_MODEL_
ID_REVISION=3.AAE
ID_SERIAL=
ID_SERIAL_
ID_SCSI_
ID_PATH=
ID_FS_VERSION=
ID_FS_TYPE=
ID_FS_USAGE=raid
ID_FS_UUID=
ID_FS_UUID_
DKD_MEDIA_
MD_LEVEL=raid1
MD_DEVICES=2
MD_UUID=
MD_UPDATE_
MD_EVENTS=160863
DKD_PRESENTATIO
DEVNAME=/dev/sda2
MAJOR=8
MINOR=2
DEVLINKS=
UDEV [1255421703.633748] add /devices/
UDEV_LOG=3
ACTION=add
DEVPATH=
SUBSYSTEM=block
DEVTYPE=partition
SEQNUM=2054
ID_TYPE=disk
ID_BUS=ata
ID_MODEL=
ID_MODEL_
ID_REVISION=3.AAE
ID_SERIAL=
ID_SERIAL_
ID_SCSI_
ID_PATH=
ID_FS_VERSION=
ID_FS_TYPE=
ID_FS_USAGE=raid
ID_FS_UUID=
ID_FS_UUID_
DKD_MEDIA_
MD_LEVEL=raid0
MD_DEVICES=2
MD_UUID=
Scott James Remnant (Canonical) (canonical-scott) wrote : | #22 |
Also
UDEV [1255421699.902552] add /devices/
UDEV_LOG=3
ACTION=add
DEVPATH=
SUBSYSTEM=block
DEVTYPE=disk
SEQNUM=2781
MD_LEVEL=raid1
MD_DEVICES=2
MD_METADATA=00.90
MD_UUID=
ID_FS_UUID=
ID_FS_UUID_
ID_FS_SEC_TYPE=ext2
ID_FS_VERSION=1.0
ID_FS_TYPE=ext3
ID_FS_USAGE=
DKD_MEDIA_
DKD_PRESENTATIO
DEVNAME=/dev/md0
MAJOR=9
MINOR=0
DEVLINKS=
UDEV [1255421701.429721] change /devices/
UDEV_LOG=3
ACTION=change
DEVPATH=
SUBSYSTEM=block
DEVTYPE=disk
SEQNUM=2807
MD_LEVEL=raid1
MD_DEVICES=2
MD_METADATA=00.90
MD_UUID=
ID_FS_UUID=
ID_FS_UUID_
ID_FS_SEC_TYPE=ext2
ID_FS_VERSION=1.0
ID_FS_TYPE=ext3
ID_FS_USAGE=
FSTAB_NAME=/dev/md1
FSTAB_DIR=/home
FSTAB_TYPE=ext3
FSTAB_OPTS=noatime
FSTAB_FREQ=0
FSTAB_PASSNO=2
DKD_MEDIA_
DKD_PRESENTATIO
DEVNAME=/dev/md1
MAJOR=9
MINOR=1
DEVLINKS=
UDEV [1255421699.949313] add /devices/
UDEV_LOG=3
ACTION=add
DEVPATH=
SUBSYSTEM=block
DEVTYPE=disk
SEQNUM=2783
MD_LEVEL=raid0
MD_DEVICES=2
MD_METADATA=00.90
MD_UUID=
ID_FS_UUID=
ID_FS_UUID_
ID_FS_TYPE=xfs
ID_FS_USAGE=
FSTAB_NAME=/dev/md2
FSTAB_DIR=/scratch
FSTAB_TYPE=xfs
FSTAB_OPTS=noatime
FSTAB_FREQ=0
FSTAB_PASSNO=2
DKD_MEDIA_
DKD_PRESENTATIO
DEVNAME=/dev/md2
MAJOR=9
MINOR=2
DEVLINKS=
Scott James Remnant (Canonical) (canonical-scott) wrote : | #23 |
robegue: your collected data does not support your assertion, all your drives are detected and your RAID arrays are assembled and active. If you have a problem, it's with mdadm
robegue (r087r70) wrote : | #24 |
>robegue: your collected data does not support your assertion, all your drives
>are detected and your RAID arrays are assembled and active. If you
>have a problem, it's with mdadm
Of course my RAID are assembled and active: as stated above I'm using the 2.6.27 kernel with a working initrd image!
The problem is that I can not use any newer kernel because I can not generate any new working initrd image!
Jyrki Pulliainen (jyrki-pulliainen) wrote : apport-collect data | #25 |
Architecture: i386
CustomUdevRuleF
DistroRelease: Ubuntu 9.10
MachineType: System manufacturer P5K PRO
NonfreeKernelMo
Package: udev 147~-6
PackageArchitec
ProcCmdLine: root=/dev/md1 ro quiet splash
ProcEnviron:
SHELL=/bin/bash
LANG=en_DK.UTF-8
LANGUAGE=
ProcVersionSign
Uname: Linux 2.6.31-
UserGroups:
dmi.bios.date: 04/18/2008
dmi.bios.vendor: American Megatrends Inc.
dmi.bios.version: 1002
dmi.board.
dmi.board.name: P5K PRO
dmi.board.vendor: ASUSTeK Computer INC.
dmi.board.version: Rev 1.xx
dmi.chassis.
dmi.chassis.type: 3
dmi.chassis.vendor: Chassis Manufacture
dmi.chassis.
dmi.modalias: dmi:bvnAmerican
dmi.product.name: P5K PRO
dmi.product.
dmi.sys.vendor: System manufacturer
Jyrki Pulliainen (jyrki-pulliainen) wrote : BootDmesg.txt | #26 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : CurrentDmesg.txt | #27 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : Dependencies.txt | #28 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : Lspci.txt | #29 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : Lsusb.txt | #30 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : ProcCpuinfo.txt | #31 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : ProcInterrupts.txt | #32 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : ProcModules.txt | #33 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : UdevDb.txt | #34 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : UdevLog.txt | #35 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : Re: udev causes raid to degrade after update to Karmic beta | #36 |
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #37 |
Just a quick update: After RC came out, the partitions are still missing
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #38 |
Final out, still no change.
However, could this be related to the Intel fake raid controller I have onboard? It's disabled in bios (the SATA mode is AHCI, not RAID), but still worth noting
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #39 |
Oh, and the motherboard chipset is Intel P35 (Asus P5K Pro)
robegue (r087r70) wrote : | #40 |
I don't think it's related to the intel controller because I'm on nvidia controller (disabled in bios) and feell the same bug.
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #41 |
Ah, problem solved: The reason is that mdadm thinks the partitions are controlled by fake raid on motherboard. Adding nodmraid to boot options solves the problem.
However, this is still an issue, as the motherboard raid is disabled on bios and the mdadm should not use dmraid then
Scott James Remnant (Canonical) (canonical-scott) wrote : | #42 |
Moving to dmraid given "nodmraid" comment
affects: | udev (Ubuntu) → dmraid (Ubuntu) |
summary: |
- udev causes raid to degrade after update to Karmic beta + raid degraded after update to Karmic beta |
Tormod Volden (tormodvolden) wrote : | #43 |
The fake raid is on the disks, not on the motherboard. Linux/dmraid can not see if you have disabled the fakeraid support in the BIOS, but it sees the fakeraid signatures on the disks. If your BIOS does not have an option to delete the fakeraid configuration from the disks, use dmraid -E.
robegue (r087r70) wrote : | #44 |
using nodmraid in boot options works.
Instead, i'm not sure how to use the dmraid -E to definitely solve the issue: is it a safe operation? how is the full command line for /dev/md0?
Jyrki Pulliainen (jyrki-pulliainen) wrote : | #45 |
Use dmraid -r to see, if your system has any dmraid controlled disks. Then, to remove the dmraid signatures, use: dmraid -r /dev/sdX -E, where /dev/sdX is a disk displayed by the dmraid -r command. A safe bet is to do a cross compare between the devices listed by dmraid -r and the devices listed in /proc/mdstat, the ones in /proc/mdstat *probably* should not be controlled by dmraid.
However, please note that removing the signatures from wrong disks might render your system unbootable or it might even result in a data loss
NaSH (lenashou) wrote : | #46 |
it seems that you're right, it's an old signature on disks. I was using some of these disks to test raid, 2 years ago..
sudo dmraid -r
/dev/sda: "sil" and "nvidia" formats discovered (using nvidia)!
/dev/sdc: nvidia, "nvidia_cbccehbc", stripe, ok, 488397166 sectors, data@ 0
/dev/sdb: nvidia, "nvidia_cbccehbc", stripe, ok, 488397166 sectors, data@ 0
/dev/sda: nvidia, "nvidia_fhjffiga", stripe, ok, 490234750 sectors, data@ 0
so i guess that signatures are on sda only ? or on sdb and sdc too ?
so, trying dmraid -r /dev/sda -E should solve the pblm
robegue (r087r70) wrote : | #47 |
$ sudo dmraid -r
/dev/sdb: nvidia, "nvidia_cbajegji", mirror, ok, 625142446 sectors, data@ 0
/dev/sda: nvidia, "nvidia_cbajegji", mirror, ok, 625142446 sectors, data@ 0
$
can I wait for a dmraid patch or MUST I dmraid -E the disks? I'm afraid to lose data since this is a production machine...
Tormod Volden (tormodvolden) wrote : | #48 |
If you have false metadata on the disks, you _should_ remove this. Meanwhile you can use the nodmraid boot option to ignore all fakeraid signatures.
Danny Wood (danwood76) wrote : | #49 |
There will not be a patch to dmraid as this is not really a bug.
This has been discussed before.
The issue is that you haven't removed the raid metadata when you destroyed the array.
Normally you can remove these without issue using dmraid -E as the metadata is usually in a region of the disk that you cannot access anyhow.
Also its worth going through the BIOS route to see if you can delete the array first, you may need to enable the onboard RAID to get into the nvidia / intel / whatever control panel to do it. This way dmraid wont delete anything that might be useful.
robegue (r087r70) wrote : | #50 |
Thanks for the replies. It's still not clear to me why if I boot with the 2.6.27-14 jaunty kernel this issue does not appear? Why with the 2.6.31 karmic kernel I need to pass the nodmraid option or even modify the disks metadata, while with previous kernels I didn't?
Danny Wood (danwood76) wrote : | #51 |
It's because dmraid now hides the root devices of a RAID set. This was implemented to fix a whole host of issues with block devices and udev. Also it makes the whole dmraid system a lot better.
This is why you are getting issues but your issues are caused because you have the RAID metadata on your disks.
Just purge the metadata and you wont get dmraid hiding your block devs anymore.
The data you loose is in the last blocks on the disk and will not be used by anything, the fact that dmraid can detect the data there means it can delete it.
Michael Nagel (nailor) wrote : | #52 |
closing task on NULL project
Changed in null: | |
status: | New → Invalid |
Eva Drud (eva-drud) wrote : | #53 |
To me it's still not really clear why the installer (as well as gparted, as it seems) hide the devices, while I can access them via Nautilus in the live system. I did not try writing, though. This confused me a lot. Is there a reason for this behaviour?
Danny Wood (danwood76) wrote : | #54 |
It hides them so that you cannot accidentally destroy your RAID array when installing with Ubiquity or using the system in general.
Changed in dmraid (Ubuntu): | |
status: | New → Invalid |
no longer affects: | null |
Change to correct Project