Hi Jon,
What exactly does it mean when you say CephFS is not "production ready"?
To me, this typically indicates a product that still has business
crippling bugs.
Exactly - it means that you could and probably will stumble upon bugs
while running CephFS. I.e. if you haven't got access to someone that can
help you when you run into these bugs, then you shouldn't use it for
critical business functions.
to accomplish with Ceph is a centralized storage space for my
Hypervisors in addition to providing storage for the virtual machines
For this you do not need CephFS at all. You can use Ceph and its RBD
(Rados Block Device).
and and any other physical machines in the cluster as they need it. The
way OpenNebula works is (assume /var/lib/one is ~) it creates virtual
Actually OpenNebula can work in other ways.
RBDs in Ceph are block devices that are backed by the Ceph Storage. A
block device is similar in function to a regular, raw harddrive - except
that is available via the network and backed by a Ceph system.
You can do with RBDs what you would do with an ordinary harddrive. For
example you can mount it, partition it and format it with a file system.
Because it is backed by Ceph you can access the RBD from any server
connected to the network Ceph is connected to.
In your case this means that you can choose to power up a virtual
machine on any of your servers, and it will be able to find and use its
virtual machine image on the Ceph storage.
Your earlier postings seemed to indicate that you wanted multiple
servers using the same image at the same time. This introduces an extra
layer of complexity - but it doesn't seem to me now that this is
something you need at all.
You can power up a virtual machine backed by Ceph storage on one server
and live migrate it to a different server. This will work just fine as
KVM will coordinate the handover between the two servers so that they
won't use the RBD at the same time in a conflicting way.
OpenNebula version 4.0 has added extra support for Ceph. It is currently
out in a release candidate. As you're just starting to build your
system, it would probably be a good idea to test with this release
candidate and upgrade to 4.0 when the final release is made.
That will save you a lot of time trying to customize OpenNebula for
Ceph, as that work has already been done by the OpenNebula team.
--
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
