Hi guys,
I'm looking to setup an open source cloud IaaS system that will work
well together with Ceph. I'm looking for a system that will handle
running KVM virtual servers with persistent storage on a number of
physical servers with a multi-tenant dashboard.
I have now tried a number of systems, but having difficulties finding
something that will work with Ceph in an optimal way. Or at least,
having difficulties finding hints on how to achieve that.
By optimal I mean:
a) To have Ceph as the only storage, so that I don't have a NFS SPoF nor
have to wait for images to be copied from server to server.
b) To run KVM with the async flush feature in 1.4.2 (or backported) and
with the librbd cache.
Any of you guys are doing this? - have hints to offer?
I have tried CloudStack, but found that it was not possible to rely
fully on Ceph storage. I learnt that it would potentially be possible
with the upcoming 4.2 release, so I tried installed CloudStack from the
development source code tree. I wasn't able to get this working because
of various bugs (to be expected when running a development version
ofcourse).
I also tried OpenNebula, but found that it was very hard to get working
on the recommended CentOS 6.4 distribution. By upgrading all sorts of
systems and manually patching parts of the system I was able to get it
"almost working". However in the end, I ended up in a dilemma where
OpenNebula needed a newer qemu version to support RBDs and that newer
qemu didn't work well with the older libvirt. On the other hand if I
upgraded libvirt, I couldn't get it to work with the older qemu versions
with backported RBD support, as the newer libvirt where setting an
auth_supported=none option that stopped it from working. It didn't seem
possible to convince OpenNebula to store a secret for Ceph with libvirt.
I have been looking at OpenStack, but by reading the documentation and
googling it seems that it is not possible to configure OpenStack to use
the librbd cache with Ceph. Could this be right?
Or is it merely the case, that you cannot configure it on a per-VM
basis, so that you have to rely on the default settings in ceph.conf?
(which wouldn't be a problem for me)
Any advice you could give would be greatly appreciated!
Thanks,
--
Jens Kristian Søgaard, Mermaid Consulting ApS,
jens@xxxxxxxxxxxxxxxxxxxx,
http://www.mermaidconsulting.com/
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
