Open Stack

On 25/06, Swogat Pradhan wrote:
> Hi Rajat,
> I was using the 1st link as a reference. As per the 2nd link The rbd
> mirroring in ceph is configured properly, we've tested by creating an image
> in ceph itself the mirroring is functioning properly. The problem we are

Hi,

If I remember correctly there are 2 types of mirroring, per-pool and
per-image.

When you say it's mirroring properly, do you mean you have enabled pool
level replication?  As in, any image created in that pool will be
automatically replicated?

If that's the case, then you have to change it, because Cinder
replication works on a per image basis.


> facing is in cinder, apparently there is a bug
> https://bugs.launchpad.net/cinder/+bug/1898728 which states that cinder


> retype (to replication enabled volume type) does not enable replication and
> when we are creating a volume directly in cinder using replication enabled
> volume type, the volume is getting created in ceph but in cinder the volume
> stays in error state (as a result we are unable to use it in instances
> primary disk)

Those are 2 different cases, creating a new volume is different from
retyping.

The bug you mention describes that the problem on retype happens because
the new type is a dictionary and not a Volume instance, so the
extra_specs attribute doesn't exist.

In your case I would say the issue is related to how the pool is
configured in the Ceph cluster.

Cheers,
Gorka.

>
> On Fri, Jun 25, 2021 at 9:57 AM Rajat Dhasmana <rdhasman at redhat.com> wrote:
>
> > Hi,
> >
> > On Thu, Jun 24, 2021 at 6:11 PM Swogat Pradhan <swogatpradhan22 at gmail.com>
> > wrote:
> >
> >> Hi,
> >> I am trying to configure cinder replication with ceph as backend and i
> >> followed this URL to configure the cinder replication:
> >> https://netapp.io/2016/10/14/cinder-replication-netapp-perfect-cheesecake-recipe/
> >> and
> >> http://www.sebastien-han.fr/blog/2017/06/19/OpenStack-Cinder-configure-replication-api-with-ceph/
> >>
> >>
> > The first link is specific to netapp driver so I'm not sure how it helps
> > with ceph replication. For the second link, did you configure two ceph
> > clusters properly and set the needed values with replication_device
> > parameter?
> >
> >
> >> i created a Volume type - REPL, with replication enabled <is> True and
> >> colume backend name parameter.
> >>
> >> But when i am trying to create a volume using the created volume type the
> >> volume is getting created in ceph but in cinder the status is showing
> >> error. and when checked log getting the following error:
> >>
> >> 2021-06-24 17:59:39.556 28472 ERROR cinder.volume.drivers.rbd
> >> [req-6e6b901e-f2dd-43a4-a089-6c734b279e17 d576181b6a444541b8ec7f37d750a4c1
> >> a9ac2eba50d84d64ac44b179bbcc9183 - default default] Error creating rbd
> >> image volume-361f35c4-a589-4653-b404-fb93d2f598cf.:
> >> cinder.exception.ReplicationError: Volume
> >> 361f35c4-a589-4653-b404-fb93d2f598cf replication error: Failed to enable
> >> image replication
> >>
> >>
> > Any operation could be failing in this method[1] on rbd side to raise the
> > above error. I suggest it's better to check if you've setup replication
> > correctly.
> >
> > [1]
> > https://github.com/openstack/cinder/blob/stable/victoria/cinder/volume/drivers/rbd.py#L802
> >
> >
> >> I tried to create a volume using volume type- default and then
> >> later changed the volume type using command 'cinder retype' to the created
> >> volume type and it changed without any issue.
> >>
> >> I am using openstack victoria .
> >>
> >> With regards,
> >> Swogat Pradhan
> >>
> >