9:27 a.m.
On Mon, Nov 23, 2015 at 3:17 AM, Kenneth Marsh <kenskyfish(a)gmail.com> wrote:
Hello all,
I do development operations for a part of a software division of a large
multinational. I'm an experienced user of VMWare and Amazon AWS, soon to be
pushed onto Azure, and I've found a common thread among all solutions - they
are all expensive enough that my budget will certainly not be approved with
them. I'm deferred to the IT part of the organisation, which operates too
slowly and inefficiently (in terms of both cost and time) for my
requirements. This is what led me to RHEV, and ultimately to oVirt. This is
a feasibility study for what may ultimately be a RHEV-based data center in a
new office, and if I succeed we will be doing more on a fixed budget by
using more RHEV and less Azure.
I spent the weekend working with oVirt and I'm very impressed. I had no idea
such a comprehensive enterprise-class solution was even available. Being a
complete newcomer, I started without a clue and after a weekend had set up a
nearly-working data centre including an oVirt hypervisor node, all on old
Dell notebooks loaned to me temporarily by our IT group. I started with RHEV
but decided to use oVirt for two reasons - one being to see what's possible
with the latest and greatest, the other because RHEV required some licensing
I've not yet purchased. Long term it'll have to be RHEV for enterprise
support reasons I'm sure many are familiar with.
There are a few things I found, from a newcomer's perspective, very unclear.
What is oVirt, vs oVirt engine, vs oVirt node, vs oVirt host. Try to find
documentation on any of these and get spammed with references to the others.
I think I've worked out that these are the collective suite of products, the
management centre, the bare-metal hypervisor, and participating member
servers, respectively.
You got it mostly right.
Did you have a look at [1]?
The engine is the process that manages the whole thing. It exposes an admin
interface (web+api), monitors the hosts and VMs, etc.
A host and node are where VMs run. In many cases these terms are
interchangeable,
although usually "node" is a host installed with the ovirt-node image, whereas
an "ovirt host" is a host installed with a "normal" OS and later
provisioned to
be used (by installing stuff on it, usually using the "New Host" wizard in the
web ui).
[1] http://www.ovirt.org/Architecture
Also adding Mikey.
Which versions of CentOS/Fedora/oVirt Node are compatible at which
oVirt
compatibility level? This would normally be addressed in the release notes.
I assume you refer to 3.6, released a few weeks ago.
The release notes page [2] does mention specific versions of fedora
and el. I agree
it could be written more clearly. The Download page [3] lists them too.
[2] http://www.ovirt.org/OVirt_3.6_Release_Notes
[3] http://www.ovirt.org/Download
It was also confusing to discover oVirt node 3.2.1 is compatible at
the 3.5
3.2.1? Where is this one referenced?
RN [2] has a section "ovirt node" with links to pre-release (has
ovirt-node-iso-3.6-0.999.201510221942) and final (still empty).
Adding Fabian.
level. The answer to this remains unclear but I'm trying to use
Fedora 22
across the board now with oVirt node 3.2.1 and this seems to be working,
although I haven't gotten a server node into a cluster yet, only oVirt
nodes.
Not sure I fully understand. ovirt node includes its own OS, and IIUC we only
ship one based on el7. Technically it's probably possible to build one with
fedora, never tried that.
Storage domains - much doco about them being needed and how to
configure
them but nothing about what they are or why they are needed. I would have
Architecture [1] above does have some text about that, and links at [4] which
has more.
Some more info:
Generally speaking, and until 3.5, storage was separated from what we call
"compute nodes" (those running VMs). You can bring your own external storage,
be it a mere linux machine serving its disks through nfs/iscsi/etc., or some
"dedicated" storage boxes (netapp/emc/etc.). You can also use gluster. The
engine also manages gluster storage. In 3.6 there was some work to allow
using the same hosts for both. Some of it was finished, some postponed
to 4.0. You can check e.g. [5][6].
[4] http://www.ovirt.org/Vdsm_Storage_Terminology
[5] http://www.ovirt.org/Features/GlusterFS-Hyperconvergence
[6] http://www.ovirt.org/Features/Self_Hosted_Engine_Hyper_Converged_Gluster_...
expected an oVirt node to be capable of both data and ISO storage
but
apparently there needs to be an NFS or iSCSI filesystem there first? And
there's local storage vs shared, another concept much talked about how to
prepare and add it but not explained why one would want to do that or what
it means.
I think with further internet combing and by trial-and-error I'm very likely
to figure it all out. I hope all goes well and implement this stuff in our
new data centre and then I'd be keen to contribute some of my own tech
writing.
You are very welcome :-)
Meanwhile, I hope to be active on this mailing list and I thank everyone in
advance for sharing their oVirt experience. For any who are looking at the
doco thanks much for the plethora of stuff out there already and I hope the
above bullet points help you understand where doco most needs more
attention. At least from the perspective of one who has just come across
oVirt.
Have a nice oVirt journey and thanks for the report!
Best regards,
--
Didi