Open Stack

On 06/09, KK CHN wrote:
> ExecutionError: Unexpected error while running command.
> Command: rbd map volume-4dcb78d5-e0ce-4bfc-95ae-36a4417c35d8 --pool volumes
> --id volumes --mon_host 10.236.223.131:6789 --mon_host 10.236.223.132:6789
> Exit code: 6
> Stdout: 'RBD image feature set mismatch. You can disable features
> unsupported by the kernel with "rbd feature disable
> volumes/volume-4dcb78d5-e0ce-4bfc-95ae-36a4417c35d8 journaling".\nIn some
> cases useful info is found in syslog - try "dmesg | tail".\n'
> Stderr: 'rbd: sysfs write failed\nrbd: map failed: (6) No such device or
> address\n'
>
> I am trying to add a 20 GB plain volume created through Horizon GUI to an
> existing Windows 2012 R  VM running with VirtIO drivers installed. But this
> volume adding results in the error as shown above,
>
> trace tells to fix it by    rbd feature disable   ..   what is the
> permanent fix for this issue ? and why it happens ?
>
> ( OpenStack Ussuri, Glance, Ceph , qemu-kvm)
>
> root at ctrl1:/home/cloud# ceph --version
> ceph version 14.2.19 (bb796b9b5bab9463106022eef406373182465d11) nautilus
> (stable)
> root at ctrl1:/home/cloud#

Hi,

Usually RBD volumes are directly attached on the hypervisor, but there
was an encryption performance issue [1] so nova and os-brick added
support to attaching these volumes to the host and then use them on the
VM.

Attaching volumes on the host uses the krbd module, whereas attaching
them to the hypervisor directly uses librbd, with the first one lagging
in terms of Ceph features.

The issue you are facing is that the images in RBD are being created
with features that are not supported by krbd.

Fixes are:

- Disable those features in your ceph configuration file (usually
  /etc/ceph/ceph.conf)

  Under [global] section add:
     rbd default features = 3

- Change "[workarounds]rbd_volume_local_attach" to False in Nova's
  config.

Cheers,
Gorka.

[1]: https://bugzilla.redhat.com/show_bug.cgi?id=1762765