Hello,

On 06/25/2014 03:03 AM, NeilBrown wrote:

On Tue, 24 Jun 2014 17:38:30 +0200 Francis Moreau [off-list ref]
wrote:

quoted

Hello,

I'm having the folloing behaviour with kernel 3.14.5 and mdadm v3.3.1.

After stopping all arrays, I still can see one of them in /sys/block/:

# cat /proc/mdstat
Personalities : [raid1]
md125 : active raid1 sdb3[1] sda3[0]
      483688448 blocks super 1.2 [2/2] [UU]
      [======>..............]  resync = 34.9% (169161280/483688448)
finish=44.0min speed=118852K/sec
      bitmap: 3/4 pages [12KB], 65536KB chunk

md126 : active raid1 sdb2[1] sda2[0]
      4038656 blocks super 1.2 [2/2] [UU]

md127 : active raid1 sdb1[1] sda1[0]
      524224 blocks super 1.0 [2/2] [UU]

unused devices: <none>

# mdadm --stop /dev/md12[567]
mdadm: stopped /dev/md125
mdadm: stopped /dev/md126
mdadm: stopped /dev/md127

# cat /proc/mdstat
Personalities : [raid1]
unused devices: <none>

# ls /sys/block/
md126  sda  sdb  sdc  sr0

# ls /sys/block/md126/md/
array_size  array_state  bitmap  chunk_size  component_size  layout
level  max_read_errors  metadata_version  new_dev  raid_disks
reshape_direction  reshape_position  resync_start  safe_mode_delay

# dmesg
....
[ 1573.715476] md125: detected capacity change from 495296970752 to 0
[ 1573.715626] md: md125 stopped.
[ 1573.715633] md: unbind<sdb3>
[ 1573.740681] md: export_rdev(sdb3)
[ 1573.740694] md: unbind<sda3>
[ 1573.754008] md: export_rdev(sda3)
[ 1573.773398] md126: detected capacity change from 4135583744 to 0
[ 1573.773403] md: md126 stopped.
[ 1573.773410] md: unbind<sdb2>
[ 1573.820652] md: export_rdev(sdb2)
[ 1573.820664] md: unbind<sda2>
[ 1573.873974] md: export_rdev(sda2)
[ 1573.889904] md127: detected capacity change from 536805376 to 0
[ 1573.889910] md: md127 stopped.
[ 1573.889917] md: unbind<sdb1>
[ 1573.913978] md: export_rdev(sdb1)
[ 1573.914033] md: unbind<sda1>
[ 1573.940627] md: export_rdev(sda1)

After waiting a couple of min, stopping again md126 worked:

[ 1835.755661] md: md126 stopped.

Is this expected ?

No overly surprising.

This is probably caused by udev, or something udev runs, opening /dev/md126
after it has been stopped.  This has the effect of creating an empty inactive
array.
e.g.

Sorry for resurecting this again, but I'm still seeing this.

Sebastian saw the same behaviour with udev 215 and it appeared that this
specific version introduced a regression which resulted in the same
behavior I described initialy.

But in my case, the version of udev used is older (204).

I tried to find out what could have opened the md device by using fuser,
but fuser reports no users.

I took a look to the udev rules which are the one shipped by mdadm 3.3.2
but nothing keep the device opened during the remove event.

Could you give me some hints here to debug this ?

Thanks