Open Stack

I guess we can dig into the details here when we get into the actual writing.  The storage issue seems significant to me --I can tolerate a bit of downtime more easily than I can tolerate losing all my data...  As I understand it, with Cinder LVM,using RAID can protect against a hard drive failure but, if you lose the Cinder LVM node itself, all the volumes stored onthat node are lost.  The only alternative I know of is to use Ceph as the storage backend for Cinder.
It may be that, for the ha-guide, we only need to explain this and link to the Ceph docs for information about configuringCeph.
 
      From: Matt Kassawara <mkassawara at gmail.com>
 To: Meg McRoberts <dreidellhasa at yahoo.com> 
Cc: "openstack-docs at lists.openstack.org" <openstack-docs at lists.openstack.org> 
 Sent: Thursday, March 12, 2015 4:33 PM
 Subject: Re: [OpenStack-docs] [ha-guide] Observations from mid-cycle ops meetup
   
Ceph definitely has a following, but I don't recall specific conversation around redundant storage. Similar to physical network redundancy, I wonder if we should simply mention storage redundancy options rather than attempt to support a particular variant... especially if all of them lean toward a vendor.


On Thu, Mar 12, 2015 at 4:58 PM, Meg McRoberts <dreidellhasa at yahoo.com> wrote:

Thanks, Matt -- this is very helpful.
What about storage?  It seems like Ceph as the storage back-end provides much better data protectionthan the alternatives.  Did you sense a movement towards adapting Ceph or are people clinging to LVM,etc?  

 
      From: Matt Kassawara <mkassawara at gmail.com>
 To: "openstack-docs at lists.openstack.org" <openstack-docs at lists.openstack.org> 
 Sent: Thursday, March 12, 2015 1:44 PM
 Subject: [OpenStack-docs] [ha-guide] Observations from mid-cycle ops meetup
   
I went to the mid-cycle ops meeting for a better idea of the HA methods implemented by operators. Most seem to agree on Galera with MariaDB/MySQL, RabbitMQ active-passive/active-active (some using a load balancer and others configuring services to use each node directly), Memcached active/active (via Oslo hash synchronization), and OpenStack APIs via hardware or HAProxy load balancer. On the other hand, networking drifted all over the place. Some people live and die by nova-net and won't move to neutron until it 100% mirrors the multi-host functionality with fixed/floating IP addresses, others use neutron but only with provider networks, most don't quite trust DVR/L3HA yet, and a few implement custom code with nova-net or neutron. For networking in the HA guide, I mostly suggest referencing the networking guide scenarios rather than suggesting a particular architecture. However, we should also mention them in the introduction somewhere because different architectures may impact the minimum number and type of nodes.
_______________________________________________
OpenStack-docs mailing list
OpenStack-docs at lists.openstack.org
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-docs


   
 



   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-docs/attachments/20150312/031f3e9e/attachment-0001.html>