Linux software RAID 1 - root filesystem becomes read-only after a fault on one disk

Question

Linux software RAID 1 locking to read-only mode

The setup:
Centos 5.2, 2x 320 GB sata drives in RAID 1.

• /dev/md0 (/dev/sda1 + /dev/sdb1) is /boot
• /dev/md1 (/dev/sda1 + /dev/sdb1) is an LVM partition which contains /, /data and swap partitions

All filesystems other than swap are ext3

We've had problem on several systems where a fault on one drive has locked the root filesystem as readonly, which obviously causes problems.

[root@myserver /]# mount | grep Root
/dev/mapper/VolGroup00-LogVolRoot on / type ext3 (rw)
[root@myserver /]# touch /foo
touch: cannot touch `/foo': Read-only file system

I can see that one of the partitions in the array is faulted:

[root@myserver /]# mdadm --detail /dev/md1
/dev/md1:
[...]
          State : clean, degraded
 Active Devices : 1
Working Devices : 1
 Failed Devices : 1
  Spare Devices : 0
[...]
    Number   Major   Minor   RaidDevice State
       0       0        0        0      removed
       1       8       18        1      active sync   /dev/sdb2
       2       8        2        -      faulty spare   /dev/sda2

Remounting as rw fails:

[root@myserver /]# mount -n -o remount /
mount: block device /dev/VolGroup00/LogVolRoot is write-protected, mounting read-only

The LVM tools give an error unless --ignorelockingfailure is used (because they can't write to /var) but show the volume group as rw:

[root@myserver /]# lvm vgdisplay
Locking type 1 initialisation failed.
[root@myserver /]# lvm pvdisplay --ignorelockingfailure
  --- Physical volume ---
  PV Name               /dev/md1
  VG Name               VolGroup00
  PV Size               279.36 GB / not usable 15.56 MB
  Allocatable           yes (but full)
  [...]

[root@myserver /]# lvm vgdisplay --ignorelockingfailure
  --- Volume group ---
  VG Name               VolGroup00
  System ID
  Format                lvm2
  Metadata Areas        1
  Metadata Sequence No  4
  VG Access             read/write
  VG Status             resizable
  [...]

[root@myserver /]# lvm lvdisplay /dev/VolGroup00/LogVolRoot --ignorelockingfailure
  --- Logical volume ---
  LV Name                /dev/VolGroup00/LogVolRoot
  VG Name                VolGroup00
  LV UUID                PGoY0f-rXqj-xH4v-WMbw-jy6I-nE04-yZD3Gx
  LV Write Access        read/write
  [...]

In this case /boot (seperate RAID meta-device) and /data (a different logical volume in the same volume group) are still writtable. From the previous occurances I know that a restart will bring the system back up with a read/write root filesystem and a properly degraded RAID array.

So, I have two questions:

1) When this occurs, how can I get the root filesystem back to read/write without a system restart?

2) What needs to be changed to stop this filesystem locking? With a RAID 1 failure on a single disk we don't want the filesystems to lockup, we want the system to keep running until we can replace the bad disk.

---

Edit: I can see this in teh dmesg output - doe sthis indicate a failure of /dev/sda, then a seperate failure on /dev/sdb that lead to the filesystem being set to read only?

sda: Current [descriptor]: sense key: Aborted Command
    Add. Sense: Recorded entity not found

Descriptor sense data with sense descriptors (in hex):
        72 0b 14 00 00 00 00 0c 00 0a 80 00 00 00 00 00
        00 03 ce 85
end_request: I/O error, dev sda, sector 249477
raid1: Disk failure on sda2, disabling device.
        Operation continuing on 1 devices
ata1: EH complete
SCSI device sda: 586072368 512-byte hdwr sectors (300069 MB)
sda: Write Protect is off
sda: Mode Sense: 00 3a 00 00
SCSI device sda: drive cache: write back
RAID1 conf printout:
 --- wd:1 rd:2
 disk 0, wo:1, o:0, dev:sda2
 disk 1, wo:0, o:1, dev:sdb2
RAID1 conf printout:
 --- wd:1 rd:2
 disk 1, wo:0, o:1, dev:sdb2
ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
ata2.00: irq_stat 0x40000001
ata2.00: cmd ea/00:00:00:00:00/00:00:00:00:00/a0 tag 0
         res 51/04:00:34:cf:f3/00:00:00:f3:40/a3 Emask 0x1 (device error)
ata2.00: status: { DRDY ERR }
ata2.00: error: { ABRT }
ata2.00: configured for UDMA/133
ata2: EH complete



sdb: Current [descriptor]: sense key: Aborted Command
    Add. Sense: Recorded entity not found

Descriptor sense data with sense descriptors (in hex):
        72 0b 14 00 00 00 00 0c 00 0a 80 00 00 00 00 00
        01 e3 5e 2d
end_request: I/O error, dev sdb, sector 31677997
Buffer I/O error on device dm-0, logical block 3933596
lost page write due to I/O error on dm-0
ata2: EH complete
SCSI device sdb: 586072368 512-byte hdwr sectors (300069 MB)
sdb: Write Protect is off
sdb: Mode Sense: 00 3a 00 00
SCSI device sdb: drive cache: write back
ata2.00: exception Emask 0x0 SAct 0x1 SErr 0x0 action 0x0
ata2.00: irq_stat 0x40000008
ata2.00: cmd 61/38:00:f5:d6:03/00:00:00:00:00/40 tag 0 ncq 28672 out
         res 41/10:00:f5:d6:03/00:00:00:00:00/40 Emask 0x481 (invalid argument) <F>
ata2.00: status: { DRDY ERR }
ata2.00: error: { IDNF }
ata2.00: configured for UDMA/133
sd 1:0:0:0: SCSI error: return code = 0x08000002
sdb: Current [descriptor]: sense key: Aborted Command
    Add. Sense: Recorded entity not found

Descriptor sense data with sense descriptors (in hex):
        72 0b 14 00 00 00 00 0c 00 0a 80 00 00 00 00 00
        00 03 d6 f5
end_request: I/O error, dev sdb, sector 251637
ata2: EH complete
SCSI device sdb: 586072368 512-byte hdwr sectors (300069 MB)
sdb: Write Protect is off
sdb: Mode Sense: 00 3a 00 00
SCSI device sdb: drive cache: write back
Aborting journal on device dm-0.
journal commit I/O error
ext3_abort called.
EXT3-fs error (device dm-0): ext3_journal_start_sb: Detected aborted journal
Remounting filesystem read-only

Solution

Your dmesg output should give you an indication as to why it's signalling failure to the PV; that shouldn't be happening. As far as getting the system writable again, kicking the VG and LV to read-only and then back to read-write works from memory, but the real resolution is getting md to stop worrying LVM unnecessarily.

Comments

[Admin]

Did you ever find a solution to your two questions? I'm currently facing a very similar problem.

[DrStalker]

Its not possible to set the logical volume to read-only because lvchange -p coesn't support the --ignorelockingfailure parameter (only -a does), and I can't find any equivilent command for read-only mode on a volume group - any other suggestions for getting this running without a restart?