Open Stack

Hi Victor,

Please see my answers inline below.

In Newton, we added support for Generic Volume Groups.  See doc below.  CGs will be migrated to Generic Volume Groups gradually.  Drivers should not implement CGs any more.  Instead, it can add CG support using Generic Volume Group interfaces.  I'm working on a dev doc to explain how to do this and will send an email to the mailing list when I'm done.  The Generic Volume Group interface is very similar to CG interface, except that the Generic Volume Group requires an additional Group Type parameter to be created.  Using Group Type, CG can be a special type of Generic Volume Group.  Please feel free to grab me on Cinder IRC if you have any questions.  My IRC handle is xyang or xyang1.

http://docs.openstack.org/admin-guide/blockstorage-groups.html

Thanks,
Xing


________________________________________
From: Victor Denisov [vdenisov at mirantis.com]
Sent: Monday, October 31, 2016 11:29 PM
To: openstack-dev at lists.openstack.org
Cc: Jason Dillaman
Subject: [openstack-dev] [cinder] consistency groups in ceph

Hi,

I'm working on consistency groups feature in ceph.
My question is about what kind of behavior does cinder expect from
storage backends.
I'm particularly interested in what happens to consistency groups
snapshots when I remove an image from the group:

Let's imagine I have a consistency group called CG. I have images in
the consistency group:
Im1, Im2, Im3, Im4.
Let's imagine we have snapshots of this consistency group:

CGSnap1
CGSnap2
CGSnap3

Snapshots of individual images in a consistency group snapshot I will call
CGSnap2Im1 - Snapshot of image 1 from consistency group snapshot 2.

Qustion 1:
If consistency group CG has 4 images: Im1, Im2, Im3, Im4.
Can CGSnap1 have more images than it already has: Im1, Im2, Im3, Im4, Im5.

Can CGSnap1 have less images than it already has: Im1, Im2, Im3.

[Xing]  Once a snapshot is taken from a CG, it can no longer be changed.  It is a point-in-time copy.  CGSnap1 cannot be modified.

Question 2:
If we remove image2 from the consistency group. Does it mean that
snapshots of this image should be removed from all the CGSnaps.

Example:
We are removing Im2.
CGSnaps look like this:

CGSnap1 - CGSnap1Im1, CGSnap1Im2, CGSnap1Im3
CGSnap2 - CGSnap2Im1, CGSnap2Im2, CGSnap2Im3, CGSnap3Im4
CGSnap3 - CGSnap3Im1, CGSnap3Im2, CGSnap3Im3, CGSnap3Im4

What happens to snapshots: CGSnap1Im2,CGSnap2Im2, CGSnap3Im2? Do we
remove them, do we keep them. Is it important what we do to them at
all?

[Xing] If your CG contains 4 volumes when you take the snapshot of the CG, the resulting CGSnap should be associated with 4 snapshots corresponding to the 4 volumes.  If you add more volumes to the CG or remove volumes from CG after CGSnap was taken, it should not affect CGSnap.  It will only affect CG snapshots that you take in the future.

Thanks,
Victor.

__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev