How to identify a failed md array

Hello,

I am wondering how to identify a failed md array.
Lets assume the following array

/dev/md0:

        Version : 1.2

Creation Time : Mon May 26 19:10:59 2014
Raid Level : raid1
Array Size : 10176 (9.94 MiB 10.42 MB)
Used Dev Size : 10176 (9.94 MiB 10.42 MB)
Raid Devices : 2
Total Devices : 2
Persistence : Superblock is persistent

Update Time : Mon May 26 19:10:59 2014
State : clean
Active Devices : 2
Working Devices : 2
Failed Devices : 0
Spare Devices : 0

Name : test:0 (local to host test) UUID : cac8fd48:44219a96:5de7e757:4e21a3e2
Events : 17

Number Major Minor RaidDevice State
0 254 0 0 active sync /dev/dm-0
1 254 1 1 active sync /dev/dm-1

with

/sys/block/md0/md/array_state:clean
/sys/block/md0/md/dev-dm-0/state:in_sync
/sys/block/md0/md/dev-dm-1/state:in_sync

and

disk0: 0 20480 linear 7:0 0
disk1: 0 20480 linear 7:1 0

If dm-0 gets changed to "disk0: 0 20480 error" and we read from the array (dd if=/dev/md0 count=1 iflag=direct of=/dev/null) the broken disk gets detected by md:

[84688.483607] md/raid1:md0: dm-0: rescheduling sector 0
[84688.483654] md/raid1:md0: redirecting sector 0 to other mirror: dm-1
[84688.483670] md: super_written gets error=-5, uptodate=0
[84688.483672] md/raid1:md0: Disk failure on dm-0, disabling device.
md/raid1:md0: Operation continuing on 1 devices.
[84688.483676] md: super_written gets error=-5, uptodate=0
[84688.494174] RAID1 conf printout:
[84688.494178]  --- wd:1 rd:2
[84688.494181]  disk 0, wo:1, o:0, dev:dm-0
[84688.494182]  disk 1, wo:0, o:1, dev:dm-1
[84688.494183] RAID1 conf printout:
[84688.494184]  --- wd:1 rd:2
[84688.494184]  disk 1, wo:0, o:1, dev:dm-1

/dev/md0:

        Version : 1.2

Creation Time : Mon May 26 19:10:59 2014
Raid Level : raid1
Array Size : 10176 (9.94 MiB 10.42 MB)
Used Dev Size : 10176 (9.94 MiB 10.42 MB)
Raid Devices : 2
Total Devices : 2
Persistence : Superblock is persistent

Update Time : Mon May 26 19:27:41 2014
State : clean, degraded
Active Devices : 1
Working Devices : 1
Failed Devices : 1
Spare Devices : 0

Name : test:0 (local to host test) UUID : cac8fd48:44219a96:5de7e757:4e21a3e2
Events : 20

Number Major Minor RaidDevice State
0 0 0 0 removed
1 254 1 1 active sync /dev/dm-1

0 254 0 - faulty /dev/dm-0

md0 : active raid1 dm-1[1] dm-0[0](F)

      10176 blocks super 1.2 [2/1] [_U]

/sys/block/md0/md/array_state:clean /sys/block/md0/md/dev-dm-0/state:faulty,write_error /sys/block/md0/md/dev-dm-1/state:in_sync /sys/block/md0/md/degraded:1

However if I also change dm-1 to "disk1: 0 20480 error" and read again there is no visible state change:

/dev/md0:

        Version : 1.2

Creation Time : Mon May 26 19:10:59 2014
Raid Level : raid1
Array Size : 10176 (9.94 MiB 10.42 MB)
Used Dev Size : 10176 (9.94 MiB 10.42 MB)
Raid Devices : 2
Total Devices : 2
Persistence : Superblock is persistent

Update Time : Mon May 26 19:27:41 2014
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
1 254 1 1 active sync /dev/dm-1

0 254 0 - faulty /dev/dm-0

md0 : active raid1 dm-1[1] dm-0[0](F)

      10176 blocks super 1.2 [2/1] [_U]

/sys/block/md0/md/array_state:clean /sys/block/md0/md/dev-dm-0/state:faulty,write_error /sys/block/md0/md/dev-dm-1/state:in_sync /sys/block/md0/md/degraded:1

On write to the array we get

[85498.660247] md: super_written gets error=-5, uptodate=0
[85498.666464] quiet_error: 268 callbacks suppressed
[85498.666470] Buffer I/O error on device md0, logical block 2528
[85498.666476] Buffer I/O error on device md0, logical block 2528
[85498.666486] Buffer I/O error on device md0, logical block 2542
[85498.666490] Buffer I/O error on device md0, logical block 2542
[85498.666496] Buffer I/O error on device md0, logical block 0
[85498.666499] Buffer I/O error on device md0, logical block 0
[85498.666508] Buffer I/O error on device md0, logical block 1
[85498.666512] Buffer I/O error on device md0, logical block 1
[85498.666518] Buffer I/O error on device md0, logical block 2543
[85498.666524] Buffer I/O error on device md0, logical block 2543
[85498.866388] md: super_written gets error=-5, uptodate=0

and the only change is

/sys/block/md0/md/dev-dm-1/state:in_sync,write_error,want_replacement

How can I identify a failed array?
array_state reports "clean", the last raid member stays "in_sync" and the value in degraded doesn't equal raid_disks.

On Mon, 26 May 2014 20:07:11 +0200 Sebastian Herbszt <[email protected]> wrote:

Quoted 3 lines

How can I identify a failed array?
array_state reports "clean", the last raid member stays "in_sync" and the value in degraded doesn't equal raid_disks.

You know the array is "failed" when you get an IO error.

When a RAID1 array gets down to just one drive remaining, it starts acting like it is just one drive.
How do you tell if is single plain ordinary drive is failed? You get an IO error. ditto with RAID1.

NeilBrown

Quoted 3 lines · NeilBrown

On Mon, 26 May 2014 20:07:11 +0200 Sebastian Herbszt <[email protected]> wrote:

[snip]

Quoted 4 lines

Quoted 3 lines

How can I identify a failed array?
array_state reports "clean", the last raid member stays "in_sync" and the value in degraded doesn't equal raid_disks.

Quoted 9 lines

You know the array is "failed" when you get an IO error.

When a RAID1 array gets down to just one drive remaining, it starts acting like it is just one drive.
How do you tell if is single plain ordinary drive is failed? You get an IO error. ditto with RAID1.

NeilBrown

Since md knows the current state I hoped it would provide the required information for an application to distinguish between degraded and failed arrays. Any reason this is not the case?
This would allow mdadm to not only report "DegradedArray" but also "FailedArray". Currently it does not generate any event when the last drive fails.