I have two HP servers running SLES 11 Sp4 and some virtual machines (kvm). The vm’s are currently running on normal pc’s. I’d like to move them to the server. The two servers have an attached SAN (via FC), also from HP.
My idea is to store the vm images and their configuration files on the SAN. As a “load balancing for poor people” i’d like to have the vm’s distributed over the two servers, manually.
I’d like to have live migration. I think i need a cluster fs for that, right ? Which one should i choose ? OCFS2, GFS2 or Ceph ? Do they need a cluster ressource manager (like pacemaker) additionally ?
Furthermore i’d like to have the possibility to snapshot the vm’s. How good is the snapshotting of virsh ?
Which of the file systems has the possibility to snapshot ?
Due to several reasons i don’t like to create a HA solution.
I have two HP servers running SLES 11 Sp4 and some virtual machines (kvm). The vm’s are currently running on normal pc’s. I’d like to move them to the server. The two servers have an attached SAN (via FC), also from HP.
My idea is to store the vm images and their configuration files on the SAN. As a “load balancing for poor people” i’d like to have the vm’s distributed over the two servers, manually.
I’d like to have live migration. I think i need a cluster fs for that, right ?[/QUOTE]
yes, if you want to migrate the VMs, these need access to the disk image from both nodes - so you’ll need concurrent access from both servers, which will either required some network file system (typically undesirable), a cluster file system or a block device accessible from both nodes.
I do have experience with OCFS2, which does work nicely with SLES 11 SP4.
Modern implementations, i.e. within SUSE Openstack Cloud, allow to use Ceph as a backing store, which will required proper rdb (RADOS block device) support on your SLES11SP4 servers, I’m not 100% positive this is available in SLES11SP4. On top, you’d need a larger Ceph infrastructure (multiple servers). Your description rather sounds like you’re heading for a small solution - skip Ceph ;), unless your corporate environment is heading that way anyhow.
My personal choice would be OCFS2. It does work without Pacemaker, though my last according test was ages ago :-[
[QUOTE=berndgsflinux;34616]Furthermore i’d like to have the possibility to snapshot the vm’s. How good is the snapshotting of virsh ?
Which of the file systems has the possibility to snapshot ?[/QUOTE]
Why do you leave it for the others ? Don’t you knwo much about it or don’t you like to answer ?
probably a bit of both… we were running mostly Xen VMs under cluster control, that doesn’t qualify me as a reliable source for a proper answer
Do you know a good administration tool (beside virsh) ? What is about ovirt ?
We moved on to Openstack Cloud Perfect administration for our VMs… and pretty good at automated orchestration, both with Openstack tools, as well as via external mechanisms (which is our primary focus)
On 10/05/2016 10:34 AM, berndgsflinux wrote:[color=blue]
jmozdzen;34618 Wrote:[color=green]
Hi Bernd,
I’ll leave that for others to answer …
J[/color]
Hi J,
thanks for the quick answer. Why do you leave it for the others ? Don’t
you knwo much about it or don’t you like to answer ?
Do you know a good administration tool (beside virsh) ? What is about
ovirt ?[/color]
qcow2 supports snapshots, and I use them a lot and love them with multiple
operating systems, distros, and SLES versions. snapshots can be taken via
virt-manager (GUI) or the command line in various ways, so logs of
flexibility as you would expect.
–
Good luck.
If you find this post helpful and are logged into the web interface,
show your appreciation and click on the star below…
Perfect administration for our VMs… and pretty good at automated orchestration, both with Openstack tools, as well as via external mechanisms (which is our primary focus)