[Openstack] DRBD storage for Openstack installations

Oleg Gelbukh ogelbukh at mirantis.com
Mon May 30 06:42:01 UTC 2011


Hello, Nelson

Sure, it will be very interesting and useful to read a word about your
solutions.

On Sat, May 28, 2011 at 12:30 AM, Nelson Nahum <nelson at zadarastorage.com>wrote:

> At Zadara Storage, we are working on a block storage system for the cloud.
> We didn't published much info yet but if somebody is interested I will be
> happy to be in a call and explain what we are doing.
>
> Nelson Nahum
> CTO
> nelson at zadarastorage.com
>
>
>
> 2011/5/27 Oleg Gelbukh <ogelbukh at mirantis.com>
>
>> Hi
>> Our approach was defined by need to combine storage and compute on the
>> same hosts.
>> Our configuration is dual-primary, so we can run nova-compute and virtual
>> servers on both nodes and have them with write access to volumes. DRBD
>> allows this mode out-of-box now, but it requires clustered file system or
>> great caution when runnning LVM on it.
>> But nova-volume must run on one node of drbd-connected pair, while the
>> second gets copy of lvm data via drbd. The tricky part is that it seems we
>> must activate volumes and volume groups on the peer node, but automation of
>> this is relatively easy.
>> For now, we are not going to share volumes outside of drbd-peers pair for
>> live migration or as attachable volumes, except some special cases like
>> migrating VMs between drbd pairs.
>> Looking forward to read couple of words on your approach.
>>
>> 2011/5/26 Diego Parrilla SantamarĂ­a <diego.parrilla.santamaria at gmail.com>
>>
>>> Hi Oleg,
>>>
>>> thank you very much for your post, it's really didactic. We are taking a
>>> different approach for HA at storage level, but I have worked formerly with
>>> DRBD and I think it's a very good choice.
>>>
>>> I'm curious about how you have deployed nova-volume nodes in your
>>> architecture. You don't specify if the two nodes of the DRBD cluster run one
>>> or two instances of nova-volume. If you run one instance probably you have
>>> implemented some kind of fault-tolerant active-passive service if the
>>> nova-volume process fails in the active node, but I would like to know if
>>> you can run an active-active two nova-volume instances on two different
>>> physical nodes on top of the DRBD shared resource.
>>>
>>> Regards
>>> Diego
>>>
>>> --
>>> Diego Parrilla
>>> <http://www.stackops.com>*CEO*
>>> *www.stackops.com | * diego.parrilla at stackops.com** | +34 649 94 43 29 |
>>> skype:diegoparrilla*
>>> * <http://www.stackops.com>
>>>
>>>
>>>
>>> On Thu, May 26, 2011 at 1:29 PM, Oleg Gelbukh <ogelbukh at mirantis.com>wrote:
>>>
>>>> Hi,
>>>> We were researching Openstack for our private cloud, and want to share
>>>> experience and get tips from community as we go on.
>>>>
>>>> We have settled on DRBD as shared storage platform for our installation.
>>>> LVM is used over the drbd device to mange logical volumes. OCFS2 file system
>>>> is created on one of volumes, mounted and set up as *image_path* and *
>>>> instance_path* in the *nova.conf*, other space is reserved for storage
>>>> volumes (managed by nova-volume).
>>>>
>>>> As a result, we have shared storage suitable for features such as live
>>>> migration and snapshots. We also have some level of fault-tolerance, with
>>>> DRBD I/O error handling, which automatically redirects I/O requests to peer
>>>> node over network in case of primary node failure. We created script<https://github.com/Mirantis/openstack-utils/blob/master/recovery_instance_by_id.py>for bootstrapping lost VMs in two crash scenarios:
>>>> * dom0 host restart/domU failure: restore VMs on the same host
>>>> * dom0 host failure: restore VMs on peer node
>>>> We are considering such pair of servers with shared storage as a basic
>>>> block for the cloud structure.
>>>>
>>>> For whom it may interest, the details of DRBD installation are here<http://mirantis.blogspot.com/2011/05/shared-storage-for-openstack-based-on.html>.
>>>> I'll be glad to answer any questions and highly appreciate feedback on this.
>>>>
>>>> Oleg S. Gelbukh,
>>>> Mirantis Inc.
>>>> www.mirantis.com
>>>>
>>>> _______________________________________________
>>>> Mailing list: https://launchpad.net/~openstack
>>>> Post to     : openstack at lists.launchpad.net
>>>> Unsubscribe : https://launchpad.net/~openstack
>>>> More help   : https://help.launchpad.net/ListHelp
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~openstack
>> Post to     : openstack at lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~openstack
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20110530/dc7bd7a7/attachment.html>


More information about the Openstack mailing list