Created attachment 1789627 [details]
multipath.conf

Hi @all,

we observed the behavior of a VM being stopped in a FC multipath environment if one (of two) paths of a direct attached LUN with SCSI passthrough goes down.

The qemu command line for the disk is:

blockdev {"driver":"host_device","filename":"/dev/mapper/36000d31005771c00000000000000005a","aio":"native","node-name":"libvirt-4-storage","cache":{"direct":true,"no-flush":false},"auto-read-only":true,"discard":"unmap"}
 -blockdev {"node-name":"libvirt-4-format","read-only":false,"cache":{"direct":true,"no-flush":false},"driver":"raw","file":"libvirt-4-storage"} -device scsi-block,bus=ua-c31a9cae-ba40-4c5c-8748-0e9407099879.0,channel=0,scsi-id=0,lun=1,share-rw=on,drive=libvirt-4-format,id=ua-7c4e45e3-e577-48e2-bb04-c463c3596bb9,werror=stop,rerror=stop

AFAIR the werror/rerror statement above determines what should happen in case of a (read|write)error.


Using no SCSI Passthrough, werror|rerror is also set to stop, but the VM is not paused. The only difference I spot is the -device scsi-hd vs. -device scsi-block:

-blockdev {"driver":"host_device","filename":"/dev/mapper/36000d31005771c000000000000000066","aio":"native","node-name":"libvirt-1-storage","cache":{"direct":true,"no-flush":false},"auto-read-only":true,"discard":"unmap"}
-blockdev {"node-name":"libvirt-1-format","read-only":false,"cache":{"direct":true,"no-flush":false},"driver":"raw","file":"libvirt-1-storage"}
-device scsi-hd,bus=ua-eff86ed8-d744-43d1-8f69-463c4974691f.0,channel=0,scsi-id=0,lun=0,device_id=3686d9c0-d674-4ddd-b801-a69d7657ea47,drive=libvirt-1-format,id=ua-3686d9c0-d674-4ddd-b801-a69d7657ea47,bootindex=2,write-cache=on,serial=3686d9c0-d674-4ddd-b801-a69d7657ea47,werror=stop,rerror=stop

The multipath is terminated on the hypervisor and the device created by the multipath device mapper is passed into the VM.

# multipath -ll
36000d31005771c00000000000000005a dm-1 COMPELNT,Compellent Vol
size=10T features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
`-+- policy='service-time 0' prio=50 status=active
  |- 14:0:0:1  sdc  8:32   active ready running
  `- 15:0:0:1  sds  65:32  active ready running

Multipath configuration is handled by vsdm, but see attached multipath.conf.