2:25 a.m.
Hello,
We are in the middle of scaling out our OpenStack management services like
nova, keystone etc. We do it with haproxy with two active/active servers
(mgmt1 & mgmt2) which are hosting the API endpoints.
Now the question regarding glance, at the moment we have a full glance
installation on mgmt1 with the data store on the local file system. What is
the best approach to scale out glance without having doubled images?
Cheers
Chris
2:59 a.m.
New subject: [Rdo-list] [Openstack] Glance Active/active and the data store
Now the question regarding glance, at the moment we have a full
glance installation on mgmt1 with the data store on the local file system.
What is the best approach to scale out glance without having doubled images?
You will need some kind of shared storage solution.
Most people either Swift or Ceph but you could also use a nfs mount.
Cheers,
Robert van Leeuwen
3:10 a.m.
I'd say using a shared storage driver:
* rbd works ok
* sheepdog (never used but got good feedback)
* swift (remember that 5GB is the limit for large objects in Swift)
* VMWare data store is also a possibility, if you use VMWare as the hypervisor.
Regards,
H.
5:34 a.m.
New subject: [Rdo-list] [Openstack] Glance Active/active and the data store
On Mon, 30 Mar 2015, Haïkel wrote:
I'd say using a shared storage driver:
* rbd works ok
* sheepdog (never used but got good feedback)
* swift (remember that 5GB is the limit for large objects in Swift)
In case it's not clear from this, Glance can store images greater than
5GB when using Swift as a backend.
(5GB is the limit for a single swift object, but Glance uses Swifts 'Dynamic Large
Object' support to store images greater than 5GB)
* VMWare data store is also a possibility, if you use VMWare as the
hypervisor.
Regards,
H.
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack(a)lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
3537
days inactive
3537
days old
3 comments
4 participants
participants (4)
-
Chris -
Haïkel -
stuart.mclaren＠hp.com -
Van Leeuwen, Robert