BFV failed because devNode doesn't exist

Deploy an instance from SCSI image failed with errors like:

Deploy of virtual machine rhel82_bfv_will_fail on host m54zvm147 failed with exception: Build of instance 4c2334e7-c2b9-4926-9986-4ff87bea5cbe was re-scheduled: zVM Cloud Connector request failed: {'overallRC': 300, 'modID': 30, 'rc': 300, 'rs': 5, 'errmsg': "Refresh bootmap fails, error code: 6 and reason: \\nERROR: devNode /dev/disk/by-path/ccw-0.0.1b13-zfcp-0x500507680b21bac6:0x0000000000000000-part1 doesn't exist. Get fcps: 1b13 1a10 and wwpns: 500507680b21bac6 500507680b21bac7 500507680b22bac6 500507680b22bac7", 'output': ''}

Following, the known 3 causes are listed:

1. The format property of the image does not match to the real format of the uploaded image file. For example, the image file uploaded is of type QCOW2 , but the type in its properties is RAW .
2. The skip_kpartx property of multipath.conf is set to value yes .
3. If the deployed image contains LVM partition, then it is possible that the DM multipath device component(eg, sda, sdb) is not filtered in the /etc/lvm/lvm.conf.

• For a mismatch of image format property:

  • Click the details of the image to check the format property from the UI.
  • Run command qemu-img info <the path of image file> to check the real format of the image file you uploaded from your local directory. If the local image file can not be found anymore, you can download it to local directory from the storage by the command.

    openstack image save --file <file_name> <image>
    

  • Re-create the image with correct format value from the UI.

• For skip_kpartx setting:

  • Check the skip_kpartx setting by running the command multipathd show config | grep skip_kpartx on the target host to which the instance was deployed, for example:

  [root@113-cmp-cb ~]# multipathd show config | grep skip_kpartx
      skip_kpartx "yes"
  

  • If the value was set to yes , change it to no by editing the /etc/multipath.conf , for example:

  defaults {
      # change skip_kpartx to no
      skip_kpartx no
      # keep other configuration items as original
      ......
  }
  

Restart multipathd to let the changes take effect by running the command systemctl restart multipathd . Finally, verify the result as follows.

[root@113-cmp-cb ~]# multipathd show config | grep skip_kpartx
skip_kpartx "no"

• For LVM filter Device Mapper multipath device component:
  • Login to the management and compute nodes, add the following configuration into the [devices] section of the /etc/lvm/lvm.conf:

  filter = [ "r|/dev/sd.*|" ]
  

  • Retry the virtual machine deployment.