[os-brick] Fix multipathd support when 'find_multipaths' is set to 'yes'

Gorka Eguileor geguileo at redhat.com
Mon Mar 18 18:56:54 UTC 2019


On 08/03, Sahid Orentino Ferdjaoui wrote:
> There is an issue in os-brick related to multipath support for iSCSI
> connector. Basically when multipathd is configured with feature
> 'find_multipaths' set to 'yes' #1815844 [0].
>
> In this situation, multipathd automatically generate multipath device
> when it discovers:
>
>  - There are at least two non-blacklisted paths with the same WWID.
>  - The user manually forces the creation of the device by specifying a
>    device with the multipath command.
>  - A path has the same WWID as a multipath device that was previously
>    created (even if that multipath device does not currently exist).
>    Whenever a multipath device is created, multipath remembers the
>    WWID of the device so that it will automatically create the device
>    again as soon as it sees a path with that WWID.
>
> This situation happens as described in bug #1815844 [0] when we attach
> a volume, detach it and then re-attach it again.
>
>   openstack volume create pure2 --size 10 --type pure
>   openstack server add volume T1 pure2
>   openstack server remove volume T1 pure2
>   openstack server add volume T1 pure2
>
> A patch has been proposed to fix the support [1].
>
> Since we have workaround when 'find_multipaths' is set to 'no'. The
> question is: do we want to fix it?
>
> If not, I suggest to document it, that this feature from multipathd
> should be disabled. Probably a side-note in the config option
> 'volume_use_multipath' can help. We could also add a log warning in
> os-brick when we detect the issue.
>

Hi,

I believe the assessment of the bug is not correct.

In all the testings I've done while refactoring the iSCSI connector
using multipathing I always used the less favorable conditions, which
include setting find_multipaths to yes, and there was never an incidence
of the multipath not being found because the multipath had already been
added before.

It is true that I never tried the Pure backend, but I have tested other
backends such as XtremIO, Kaminario, QNAP, and 3PAR.

Even though I don't have access to a Pure storage, I'll try to do a
proper RCA and update the bug and the patch.

Cheers,
Gorka.

>
> Thanks,
> s.
>
> [0] https://bugs.launchpad.net/charm-nova-compute/+bug/1815844
> [1] https://review.openstack.org/#/c/638639/
> [2] https://docs.openstack.org/nova/rocky/configuration/config.html#libvirt.volume_use_multipath
>



More information about the openstack-discuss mailing list