Open Stack

Hello. 
We have an Openstack environment based on Redhat 6.4 + Havana release. 
We are working with bootable volumes for instances which are allocated on 
FC attached V7000. 
On the host dm-multipath is configured. 

Once a new bootable volume is created on OpenStack, the active path for 
some reason is removed and the following messages appear in 
/var/log/messages file:

May 20 14:54:12 node21144 kernel: scsi 0:0:0:4: Direct-Access     IBM 2145 
            0000 PQ: 0 ANSI: 6
May 20 14:54:12 node21144 kernel: sd 0:0:0:4: [sdj] 52428800 512-byte 
logical blocks: (26.8 GB/25.0 GiB)
May 20 14:54:12 node21144 kernel: sd 0:0:0:4: [sdj] Write Protect is off
May 20 14:54:12 node21144 kernel: sd 0:0:0:4: [sdj] Write cache: disabled, 
read cache: enabled, supports DPO and FUA
May 20 14:54:12 node21144 kernel: sd 0:0:0:4: Attached scsi generic sg11 
type 0
May 20 14:54:12 node21144 kernel: sdj: unknown partition table
May 20 14:54:12 node21144 kernel: sd 0:0:0:4: [sdj] Attached SCSI disk
May 20 14:54:12 node21144 multipathd: sdj: add path (uevent)
May 20 14:54:12 node21144 kernel: scsi 1:0:1:4: Direct-Access     IBM 2145 
            0000 PQ: 0 ANSI: 6
May 20 14:54:12 node21144 kernel: sd 1:0:1:4: Attached scsi generic sg12 
type 0
May 20 14:54:12 node21144 kernel: sd 1:0:1:4: [sdk] 52428800 512-byte 
logical blocks: (26.8 GB/25.0 GiB)
May 20 14:54:12 node21144 kernel: sd 1:0:1:4: [sdk] Write Protect is off
May 20 14:54:12 node21144 kernel: sd 1:0:1:4: [sdk] Write cache: disabled, 
read cache: enabled, supports DPO and FUA
May 20 14:54:12 node21144 kernel: sdk: unknown partition table
May 20 14:54:12 node21144 kernel: sd 1:0:1:4: [sdk] Attached SCSI disk
May 20 14:54:12 node21144 multipathd: sdk: add path (uevent)
May 20 14:54:12 node21144 multipathd: mpathz: load table [0 52428800 
multipath 1 queue_if_no_path 0 2 1 round-robin 0 1 1 8:144 1 round-robin 0 
1 1 8:160 1]
May 20 14:54:12 node21144 multipathd: mpathz: event checker started
May 20 14:54:12 node21144 multipathd: sdk path added to devmap mpathz
May 20 14:55:54 node21144 multipathd: sdj: remove path (uevent)
May 20 14:55:54 node21144 multipathd: mpathz: load table [0 52428800 
multipath 1 queue_if_no_path 0 1 1 round-robin 0 1 1 8:160 1]
May 20 14:55:54 node21144 multipathd: sdj: path removed from map mpathz
May 20 14:55:54 node21144 kernel: sd 1:0:1:4: [sdk] Result: 
hostbyte=DID_OK driverbyte=DRIVER_SENSE
May 20 14:55:54 node21144 kernel: sd 1:0:1:4: [sdk] Sense Key : Illegal 
Request [current] 
May 20 14:55:54 node21144 kernel: Info fld=0x0
May 20 14:55:54 node21144 kernel: sd 1:0:1:4: [sdk] Add. Sense: Logical 
unit not supported
May 20 14:55:54 node21144 kernel: sd 1:0:1:4: [sdk] CDB: Read(10): 28 00 
03 1f ff 80 00 00 08 00
May 20 14:55:54 node21144 kernel: device-mapper: multipath: Failing path 
8:160.
May 20 14:55:54 node21144 multipathd: 8:160: mark as failed
May 20 14:55:54 node21144 multipathd: mpathz: Entering recovery mode: 
max_retries=5
May 20 14:55:54 node21144 multipathd: mpathz: remaining active paths: 0

So the volume creation fails and new instance can't be provisioned. 

We checked multipath functionality regardless to openstack :
- V7000 volume was created manually and mapped to the compute node 
- Multipath device was created properly
- The new volume was mounted to the new VM and operation system was 
successfully installed

Our conclusion is that the problem is not related to multipath 
configuration but to openstack functionality with multipath. 
Please advise. 

Thank you. 

Alexandra Kisin
Servers & Network group, IBM R&D Labs in Israel
Unix & Virtualization Team
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20140520/7f074f97/attachment.html>