Hi Murali,
Thanks for answering. I think the issues you raised indeed make sense, and
important ones.
We need to provide backup both for:
1. Volumes
2. VM instances (VM image, VM metadata, and attached volumes)
While the Cinder-backup handles (1), and is a very mature service, it does
not provide (2), and for Cinder-backup it does not make sense to handle (2)
as well.
Backup of VMs (as a package) is beyond the scope of Cinder, which implies
that indeed something beyond Cinder should take this task.
I think this can be done by having Nova orchestrate or assist the backup,
either of volumes or VMs.
I think that from a backup perspective, there is also a need for
"consistency groups" - the set of entities (volumes) that are considered a
single logical unit and should be backup together.
This logical consistency group could be larger than a VM, but a VM is a
good starting point.
In any case, we should adopt the "off-load" approach:
1. Handle application consistency issues using Nova as it manages the VMs.
Add to Nova functionality to support live and consistent backup - including
orchestrating volume backup using Cinder
2. Have Cinder do the volume "backup", and Cinder then can delegate the
task to the Storage/hypervisor or anyone else who provide a backup driver
While a new project is a neat package that addresses the issues, but does
it worth the work?
OpenStack projects are complex, and successful projects require a lot of
work and long-term maintenance, which is the real pain for open source
projects, as the team tend to be very dynamic.
My two cents is to adopt the "nova-networking" and "nova-volume" approach,
try to extend the current work within Nova and Cinder, and if we find out
it does not make sense anymore, explain the issues, and split the work to a
new project.
This way it, if a backup project is indeed needed, you already have the
community to support the effort, and you already have a mature solution.
Regards,
__________________________________________
Ronen I. Kat
Storage Research
IBM Research - Haifa
Phone: +972.3.7689493
Email: ronen...@il.ibm.com
From: Caitlin Bestler <caitlin.best...@nexenta.com>
To: Murali Balcha <murali.bal...@triliodata.com>,
Cc: OpenStack Development Mailing List
<openstack-dev@lists.openstack.org>
Date: 31/08/2013 07:25 AM
Subject: Re: [openstack-dev] Proposal for Raksha, a Data Protection As a
Service project
On 8/30/2013 12:49 PM, Murali Balcha wrote:
> Hi Caitlin,
> Did you get a chance to look at the wiki? It describes the raksha
functionality in detail.
> It includes more than volume backup. It includes vm images, all
> volumes and network configurations associated with vms and it
> supports incremental backups too. Volume backup is essential for
> implementing backup solution but not necessarily sufficient.
Cinder already allows backing volumes up to Swift, and in fact allows
incremental backups.
Any code you write will not back up a vendor's volume more efficiently
than the vendor's code itself can.
The vendor's knowledge of how the data is stored is probably sufficient,
but in this case a vendor has a far more powerful advantage. The vendor
can transfer the volume directly to the Swift server. Your service,
since it running on a compute node rather than the vendor's box, will
first have to fetch the content and *then* send it to Swift.
That's twice as much network traffic. This is not trivial when volumes
are big, which they tend to be.
If this service is implemented, customer who are using vendor backends
such as NexentaStor, Netapp or CEPH will see their performance drop.
That will clearly be unacceptable. New featqures are not allowed to
trash existing performance, especially when they are not actually
providing any new service to customers who already have volume backends
with these features.
You would need to have a proposal to work with the existing Cinder
backend Volume Drivers that in no way removed any option vendors have
currently to optimize performance.
Doing that in a new project, rather than within Cinder, can only make
life harder on the vendors and discourage participation in OpenStack.
I believe all of the features you are looking at can be accomodated by
taskflows using the existing Volume Driver feature (as evolving) in
Cinder. A new project is not justified, and it will risk creating a
major performance regression for some customers.
_______________________________________________
OpenStack-dev mailing list
OpenStack-dev@lists.openstack.org
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
_______________________________________________
OpenStack-dev mailing list
OpenStack-dev@lists.openstack.org
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
