On 11/29/2013 03:35 AM, tristano82(a)libero.it wrote:
Hello everybody,
i'm successful using ovirt with 16 physical node, in a FC cluster with a very
BIG dell compellent (and so expensive) enterprise storage ;)
I'm researching a new architecture for a new cluster, and i want to
understand
more better the glusterFS integration in ovirt.
GlusterFS is integrated with oVirt in two ways:
1. Use oVirt to manage gluster storage configuration options.
2. Use GlusterFS as a storage domain.
As i understand you have to install a normal physical node, with also the
glusterFS package… right ? after that you have to create a new cluster in
ovirt
, new datacenter and put in this new node. in that datacenter you can create
a
new data domain ( glusterFS ) that reside on that host. right ?
A cluster in oVirt can be configured to behave in 3 ways:
(i) virtualization only
(ii) Gluster storage only
(iii) virtualization + gluster storage
You need (ii) or (iii) to provide oVirt with the ability to perform
gluster storage configuration and management. You can also configure
gluster using its CLI and use (i) or (iii) for a glusterfs storage
domain. You can also have two clusters for (i), (ii) and manage both
using oVirt.
There are two ways in which GlusterFS can be used as a storage domain:
a) Use gluster native/fuse access with POSIXFS
b) Use the gluster native storage domain to bypass fuse (with libgfapi).
We are currently addressing an issue in libvirt
(
https://bugzilla.redhat.com/show_bug.cgi?id=1017289) to enable snapshot
support with libgfapi. Once this is addressed, we will have libgfapi
support in the native storage domain. Till then, fuse would be used with
native storage domain. You can find more details about native storage
domain here:
http://www.ovirt.org/Features/GlusterFS_Storage_Domain
and… after that ? ok i have 1 node that are also my " storage ", and if i
want
to all more compute node ? … every new compute node are a new " brick " for
glusterFS so i can expand/redundant the first one ?
If you are having separate compute/virtualization and storage clusters,
you would not be required to create bricks on your compute nodes.
However if you are using a single cluster for both compute and storage
(as in iii above), you need not necessarily have bricks on all compute
nodes.
HTH,
Vijay
i don't have the architecture very clear in my mind, and the documentation
don't clarify the final architecture for this type of usage.