[Openstack] [OpenStack] [Disaster Recovery] How can you do it?

Marzi, Fausto fausto.marzi at hp.com
Tue Jul 7 11:28:34 UTC 2015


Hi Giuseppe, Avishay, all,

In the Company where I'm currently working on, there are 4 Engineers full time dedicated to Freezer.

Your consideration Avishay make totally sense. There are also other points that probably needs to be taken in consideration for DR as:

- Backup and restore synchronized across multiple services:
This is important, as for instance Nova depends on many other OS service (i.e. cinder, glance, etc).

- When users execute a backup of a VM, it is also expected that the Cinder Volumes, the images and the users/tenants can also be point-in-time restored

- It is important to be able to restore data, even if any one of the OS Services is down (i.e. Keystone or Swift are not available).

- The orchestration between Nova and Cinder to execute full VM + Cinder backups is important

- Have a unified API and Web UI in Horizon is also a nice to have

These and more are the challenges that in the Freezer Team we are solving.

Being totally honest, Freezer has been used to backup many services in a reasonably big OpenStack instance for a year now.
However, it's only 6 months there's a dedicated Team working on it, so many things still needs to be done.

For any real time interaction, we are in #openstack-freezer on Freenode , so feel free to say Hi if you are around : )

Contributors are most welcome.

Thanks,
Fausto

-------- Forwarded Message --------
Subject:        Re: [Openstack] [OpenStack] [Disaster Recovery] How can you do
it?
Date:   Tue, 7 Jul 2015 11:44:17 +0200
From:   Giuseppe Galeota <giuseppegaleota at gmail.com>
To:     Avishay Traeger <avishay at stratoscale.com>
CC:     openstack at lists.openstack.org <openstack at lists.openstack.org>



Thank you all.
What can you tell me about the following projects:

- https://wiki.openstack.org/wiki/Raksha
- https://github.com/stackforge/freezer

Will they become OpenStack project?

Thanks,
Giuseppe

2015-07-06 21:33 GMT+02:00 Avishay Traeger <avishay at stratoscale.com
<mailto:avishay at stratoscale.com>>:

    As David mentioned, there is a spectrum of possible solutions for
    disaster recovery.  The solution, as always, depends on the
    requirements.  In disaster recovery, those requirements are
    expressed as:
    1. RPO - Recovery Point Objective - "If a disaster strikes, how far
    back will it set me?"
    2. RTO - Recovery Time Objective - "If a disaster strikes, how long
    until my workload is up and running?"

    The answers to both are independent and can range from "Not at all"
    to minutes or hours or days.

    A solution with high RPO and RTO is backup.  Once in a while
    transfer your images, snapshots of your VMs and volumes, and your
    SQL DB to another site.  If disaster strikes you can manually run
    your workloads from the last backup, and you have your metadata
    backed up too (keystone users/tenants, nova flavors, etc.).

    A solution with lower RPO and RTO is replication. Work is being done
    in Cinder to enable continuous volume replication between storage
    backends.  You can put your Glance images in multiple locations or
    in multi-site Swift, and replicate your DB as well.  Ideally you
    will be able to get to some consistent state between your data and
    metadata, which is not easy.  For failover, you can use Heat or some
    other orchestration tool.

    There are solutions with zero RPO and RTO, where workloads run
    active/active in multiple sites.  There are also other solutions in
    the spectrum.

    In short, the building blocks are starting to appear, but a full
    solution is not yet available.


    On Mon, Jul 6, 2015 at 8:08 PM, David Medberry
    <openstack at medberry.net <mailto:openstack at medberry.net>> wrote:

        Yes.

        http://docs.openstack.org/openstack-ops/content/snapshots.html
        API Image Create detailed here:

        Nova:
        http://developer.openstack.org/api-ref-compute-v2.1.html
        It is a snapshot of the instance's image (not of the instance's
        in memory state).

        Cinder:
        http://developer.openstack.org/api-ref-blockstorage-v2.html

        http://lmgtfy.com/?q=nova+image+create+api
        http://lmgtfy.com/?q=cinder+snapshot+api


        On Mon, Jul 6, 2015 at 11:02 AM, Giuseppe Galeota
        <giuseppegaleota at gmail.com <mailto:giuseppegaleota at gmail.com>>
        wrote:

            Thank you David,all,

            I would like to know if are there Nova/Cinder Rest APIs to
            execute VM/volume snapshot?

            Thank you all,
            Giuseppe

            2015-07-06 18:40 GMT+02:00 David Medberry
            <openstack at medberry.net <mailto:openstack at medberry.net>>:

                Giuseppe,

                A great deal of Disaster Recovery planning relies on you
                / your organization to define what they mean by DR
                planning. Yes, Nova snaps (instance snaps) and Cinder
                snaps (volume snaps) can be used in DR. Additionally,
                depending on your architecture you can utilize Swift (if
                it is geo distributed, it gives you some measure of DR
                resilience.)

                I'm not aware of a howto or manual on this topic though
                as it varies so much based on arch and definition it's
                not effective to make a generalized recommendation.

                On Mon, Jul 6, 2015 at 10:02 AM, Giuseppe Galeota
                <giuseppegaleota at gmail.com
                <mailto:giuseppegaleota at gmail.com>> wrote:

                    Dear all,
                    how can I implement a Disaster Recovery plan in
                    OpenStack? Is there some project?

                    Can I use the Nova/Cinder APIs to create snapshot
                    and upload them somewhere?

                    Thanks,
                    Giuseppe

                    _______________________________________________
                    Mailing list:
                    http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
                    Post to     : openstack at lists.openstack.org
                    <mailto:openstack at lists.openstack.org>
                    Unsubscribe :
                    http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack





        _______________________________________________
        Mailing list:
        http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
        Post to     : openstack at lists.openstack.org
        <mailto:openstack at lists.openstack.org>
        Unsubscribe :
        http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack




    --
    *Avishay Traeger*
    /Storage R&D/

    Mobile:+972 54 447 1475
    E-mail: avishay at stratoscale.com <mailto:avishay at stratoscale.com>



    Web <http://www.stratoscale.com/> | Blog
    <http://www.stratoscale.com/blog/> | Twitter
    <https://twitter.com/Stratoscale> | Google+
    <https://plus.google.com/u/1/b/108421603458396133912/108421603458396133912/posts> |
    Linkedin <https://www.linkedin.com/company/stratoscale>








More information about the Openstack mailing list