multipath discovery failed during install due to md arrays locking individual paths

[Mauricio Faria de Oliveira]

Similarly to bug 1549504 (for LVM), but this one with LVM on top of md arrays.

During a Xenial installation, some of the multipath devices were not discovered, as individual paths were locked by MD arrays.

The fix is to deactivate MD arrays before multipath discovery (in hw-detect), and activate multipath partitions during partman/init.d (in partman-multipath), so that MD arrays can be re-activated on top of the multipath devices (slightly afterward on partman/init.d/25md-devices).

The fix also requires an extra udev rule in the mdadm assembly rules, because the device-mapper devices have no 'add' event (used by the rule for assembly), but only a 'change' event. The rule is required for the initramfs as well (automatically installed w/ the mdadm install triggers), otherwise the rootfs won't be found.

Failure:

~ # multipath -v5
<...>
Feb 22 14:46:02 | mpathb: addmap [0 1115734016 multipath 1 queue_if_no_path 1 alua 2 1 round-robin 0 1 1 8:16 1 round-robin 0 1 1 8:128 1]
Feb 22 14:46:02 | libdevmapper: ioctl/libdm-iface.c(1876): device-mapper: reload ioctl on mpathb failed: Device or resource busy
Feb 22 14:46:02 | mpathb: domap (0) failure for create/reload map
<...>
Feb 22 14:46:02 | mpathc: addmap [0 1115734016 multipath 1 queue_if_no_path 1 alua 2 1 round-robin 0 1 1 8:32 1 round-robin 0 1 1 8:144 1]
Feb 22 14:46:02 | libdevmapper: ioctl/libdm-iface.c(1876): device-mapper: reload ioctl on mpathc failed: Device or resource busy
Feb 22 14:46:02 | mpathc: domap (0) failure for create/reload map
<...>
Feb 22 14:46:02 | mpathd: addmap [0 1115734016 multipath 1 queue_if_no_path 1 alua 2 1 round-robin 0 1 1 8:48 1 round-robin 0 1 1 8:160 1]
Feb 22 14:46:02 | libdevmapper: ioctl/libdm-iface.c(1876): device-mapper: reload ioctl on mpathd failed: Device or resource busy
Feb 22 14:46:02 | mpathd: domap (0) failure for create/reload map
<...>
Feb 22 14:46:02 | mpathe: addmap [0 1115734016 multipath 1 queue_if_no_path 1 alua 2 1 round-robin 0 1 1 8:64 1 round-robin 0 1 1 8:176 1]
Feb 22 14:46:02 | libdevmapper: ioctl/libdm-iface.c(1876): device-mapper: reload ioctl on mpathe failed: Device or resource busy
Feb 22 14:46:02 | mpathe: domap (0) failure for create/reload map
<...>
Feb 22 14:46:02 | mpathf: addmap [0 1115734016 multipath 1 queue_if_no_path 1 alua 2 1 round-robin 0 1 1 8:80 1 round-robin 0 1 1 8:192 1]
Feb 22 14:46:02 | libdevmapper: ioctl/libdm-iface.c(1876): device-mapper: reload ioctl on mpathf failed: Device or resource busy
Feb 22 14:46:02 | mpathf: domap (0) failure for create/reload map

~ # dmsetup table | sort
mpatha: 0 1115455488 multipath 0 1 alua 2 1 round-robin 0 1 1 8:0 1 round-robin 0 1 1 8:112 1
mpathg: 0 1115734016 multipath 0 1 alua 2 1 round-robin 0 1 1 8:96 1 round-robin 0 1 1 8:208 1
pkvmci845--vg-root: 0 5351333888 linear 259:2 5555
pkvmci845--vg-swap_1: 0 225468416 linear 259:2 5351339443

~ # lvm pvdisplay | grep Name
  PV Name /dev/md0p3
  VG Name pkvmci845-vg

~ # grep ^md /proc/mdstat
md0 : active raid0 sdm1[4] sdl1[3] sdk1[2] sdj1[1] sdi2[0]

After the patch, the md array is on top of the multipath devices, and show up as a LVM PV correctly.

~ # grep ^md /proc/mdstat
md0 : active raid0 dm-9[1] dm-10[2] dm-8[0] dm-11[3] dm-12[4]

~ # lvm vgscan
  Reading all physical volumes. This may take a while...
  Found volume group "pkvmci845-vg" using metadata type lvm2
~ # lvm vgchange -ay
  2 logical volume(s) in volume group "pkvmci845-vg" now active
~ # lvm pvdisplay | grep Name
  PV Name /dev/md0p3
  VG Name pkvmci845-vg

[Mauricio Faria de Oliveira (mfo)]

Hi @mathieu-tl,
Can you check this bug/patches, please?
Thanks!

[Ubuntu Foundations Team Bug Bot (crichton)]

The attachment "hw-detect_3_mpathmdadm.debdiff" seems to be a debdiff. The ubuntu-sponsors team has been subscribed to the bug report so that they can review and hopefully sponsor the debdiff. If the attachment isn't a patch, please remove the "patch" flag from the attachment, remove the "patch" tag, and if you are member of the ~ubuntu-sponsors, unsubscribe the team.

[This is an automated message performed by a Launchpad user owned by ~brian-murray, for any issue please contact him.]

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2016-03-14 16:20 EDT-------
.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2016-03-21 09:49 EDT-------
Hi @mathieu-tl,

Can you confirm the milestone for this is 16.04?

Thank you!

[Mathieu Trudel-Lapierre (cyphermox)]

Yes, the milestone for this is 16.04; I'm finishing some smoketesting on these this week prior to uploading the whole thing.

[Launchpad Janitor (janitor)]

This bug was fixed in the package hw-detect - 1.114ubuntu4

---------------
hw-detect (1.114ubuntu4) xenial; urgency=medium

  [ Mauricio Faria de Oliveira ]
  * disk-detect.sh: multipath: reload udev rules before re-add devices
    to guarantee all devices may get new udev properties (LP: #1549456)
  * disk-detect.sh: multipath: deactivate LVM volumes before multipath
    discovery to unlock individual paths (LP: #1549504)
  * disk-detect.sh: multipath: deactivate md arrays before multipath
    discovery to unlock individual paths (LP: #1549506)

 -- Mathieu Trudel-Lapierre <email address hidden> Mon, 21 Mar 2016 14:24:03 -0400

[Mauricio Faria de Oliveira (mfo)]

@mathieu-tl,

Do you plan to upload the remaining patch for mdadm as well?

Thanks

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2016-03-21 20:40 EDT-------
Thanks, @mathieu-tl.