[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ale] Small Clusters for VMs

On Fri, 2016-10-28 at 10:49 -0400, DJ-Pfulio wrote:
> Thanks for responding.
> Sheepdog is the storage backend. This is the way cloud stuff works on
> the cheap. Not a NAS.??It is distributed storage with a minimal
> > redundancy set (I'm planning 3 copies).??Sheepdog only works with
> according to research, which is fine.
> > Sure, I could setup a separate storage NAS (I'd use AoE for this),
> that isn't needed. I already have multiple NFS servers, but don't use
> > them for hosting VMs today. They are used for data volumes, not
> ???????>> Opinions follow (danger if you love what I don't) <<
> Won't be using oVirt (really RHEL only and seems to be 50+ different
> > F/LOSS projects in 500 different languages [I exaggerate] ) or
> > (bad taste after running it 4 yrs).??I've never regretted switching
> ESX/ESXi and Xen to KVM, not once.

Ovirt is only 49 projects and 127 languages! Really!

Ovirt is just the web gui front end (pile of java) with a mostly python
backend that runs KVM and some custom daemons to keep track of what is
running and where. It is most certainly geared towards RHEL/CentOS.
That may be an irritant to some. I've found the tool chain to
JustWork(tm).?I need VMs to run with minimal effort on my part as I
have no time to fight the complexity. I've hacked scripts to do
coolness with KVM but found Ovirt did more than I could code up with
the time I have. It really is a GPL replacement for VMWARE Vsphere.
And won't be dedicating the entire machines just to being storage or VM
hosts, so proxmox clusters aren't an option.  The migration from plain
VMs into sheepdog appears pretty straight forward (at least on youtube).

One thing I like with Ovirt is I can install the host node code on a
full CentOS install or use the hypervisor version and dedicate a node
entirely. I've used both and found them to be well suited for keeping
VMs running. If there is an issue with a node, I have a full toolchain
to work with. I don't use the hypervisor in production.

A major issue for my use is the need to have certain VM up and running
at all times. Ovirt provides a process to migrate a VM to an alternate
host if it (host or VM) goes down. The only "gotcha" of that is the
migration hosts must provide the same cpu capabilities so no mixing of
AMD and Intel without setting the VMs to be i686.
Just doing research today. Need to sleep on it. Probably won't try
anything until Sunday night.

Download CentOS 7.2
Install VM host version
yum install epel-release
Follow direction here:?https://www.ovirt.org/release/4.0.4/
starting with:
yum install?http://resources.ovirt.org/pub/yum-repo/ovirt-release40.rpm

Be aware that when docs refer to NFS mounts, the server for that can be
one of the nodes that has drive space. ISO space is where <duh> iso
images are kept for installations. I have one win10 VM running now for
a DBA with specialty tool needs.

> > 

On 10/28/2016 10:23 AM, Beddingfield, Allen wrote:
> > > > > > > > > > > > > > > > > > > > > > 
Will you have shared storage available (shared LUN or high performance NFS for the virtual hard drives that all hosts can access?)
If so, the easiest free out of the box setup is XenServer or oVirt.  I'm familiar with XenServer, but there are some oVirt fans on here, I know.

Allen Beddingfield
Systems Engineer
Office of Information Technology
The University of Alabama
Office 205-348-2251
allen at ua.edu

On 10/28/16, 9:17 AM, "ale-bounces at ale.org on behalf of DJ-Pfulio" <ale-bounces at ale.org on behalf of DJPfulio at jdpfu.com> wrote:

    I'm a little behind the times.  Looking to run a small cluster of VM
    hosts, just 2-5 physical nodes.
    Reading implies it is pretty easy with 2-5 nodes using a mix of
    sheepdog, corosync and pacemaker running on qemu-kvm VM hosts.
    Is that true?  Any advice from people who've done this already?
    So, is this where you'd start for small home/biz redundant VM cluster?
    I've never done clustering on Linux, just Unix with those expensive
    commercial tools and that was many years ago.
    In related news - Fry's has a Core i3-6100 CPU for $88 today with their
    emailed codes.  That CPU is almost 2x faster than a first gen Core
    i5-750 desktop CPU. Clustering for data redundancy at home really is
    possible with just 2 desktop systems these days. This can be used with
    or without RAID (any sort).
    Ale mailing list
    Ale at ale.org
    See JOBS, ANNOUNCE and SCHOOLS lists at

Ale mailing list
Ale at ale.org

James P. Kinney III

Every time you stop a school, you will have to build a jail. What you
gain at one end you lose at the other. It's like feeding a dog on his
own tail. It won't fatten the dog.
- Speech 11/23/1900 Mark Twain

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ale.org/pipermail/ale/attachments/20161028/e28c05c5/attachment.html>