[ovirt-users] Ovirt nested in VMware
Gianluca Cecchi
gianluca.cecchi at gmail.com
Fri May 12 13:36:20 UTC 2017
On Fri, May 12, 2017 at 1:06 PM, Michal Skrivanek <
michal.skrivanek at redhat.com> wrote:
>
> > On 11 May 2017, at 19:52, Mark Duggan <mduggan at gmail.com> wrote:
> >
> > Hi everyone,
> >
> > From reading through the mailing list, it does appear that it's possible
> to have the ovirt nodes/hosts be VMware virtual machines, once I enable the
> appropriate settings on the VMware side. All seems to have gone well, I can
> see the hosts in the ovirt interface, but when I attempt to create and
> start a VM it never gets past printing the SeaBios version and the machine
> UUID to the screen/console. It doesn't appear to try to boot from the hard
> disk or an ISO that I've attached.
> >
> > Has anyone else encountered similar behaviour?
>
> I wouldn’t think you can even get that far.
> It may work with full emulation (non-kvm) but we kind of enforce it in
> oVirt so some changes are likely needed.
> Of course even if you succeed it’s going to be hopelessly slow. (or maybe
> it is indeed working and just runs very slow)
>
> Nested on a KVM hypervisor runs ok
>
> Thanks,
> michal
>
>
In the past I was able to get an Openstack Icehouse environment running
inside vSphere 5.x for a POC (on poweful physical servers) and performance
of nested VMs inside the virtual compute nodes was acceptable.
More recently I configured a standalone ESXi server 6.0 U2 on a Nuc6 with
32Gb of ram and 2 ssd disks and on it I have now running 2 kinds of
environments (just verified they are still on after some months I abandoned
them to their destiny... ;-)
1) an ESXi VM acting as a single oVirt host (4.1.1 final or pre, I don't
remember) with self hosted engine (that itself becomes an L2 VM) and also
another VM (CentOS 6.8)
See here a screenshot of the web admin gui with a spice console open after
connecting to the engine:
https://drive.google.com/file/d/0BwoPbcrMv8mvanpTUnFuZ2FURms/view?usp=sharing
2) a virtual oVirt gluster environment based on 4.0.5 with 3 Virtual Hosts
(with one as arbiter node if I remember correctly)
On this second environment I have ovirt01, virt02 and ovirt03 VMs:
[root at ovirt02 ~]# hosted-engine --vm-status
--== Host 1 status ==--
Status up-to-date : True
Hostname : ovirt01.localdomain.local
Host ID : 1
Engine status : {"reason": "vm not running on this
host", "health": "bad", "vm": "down", "detail": "unknown"}
Score : 3042
stopped : False
Local maintenance : False
crc32 : 2041d7b6
Host timestamp : 15340856
Extra metadata (valid at timestamp):
metadata_parse_version=1
metadata_feature_version=1
timestamp=15340856 (Fri May 12 14:59:17 2017)
host-id=1
score=3042
maintenance=False
state=EngineDown
stopped=False
--== Host 2 status ==--
Status up-to-date : True
Hostname : 192.168.150.103
Host ID : 2
Engine status : {"health": "good", "vm": "up",
"detail": "up"}
Score : 3400
stopped : False
Local maintenance : False
crc32 : 27a80001
Host timestamp : 15340760
Extra metadata (valid at timestamp):
metadata_parse_version=1
metadata_feature_version=1
timestamp=15340760 (Fri May 12 14:59:11 2017)
host-id=2
score=3400
maintenance=False
state=EngineUp
stopped=False
--== Host 3 status ==--
Status up-to-date : True
Hostname : ovirt03.localdomain.local
Host ID : 3
Engine status : {"reason": "vm not running on this
host", "health": "bad", "vm": "down", "detail": "unknown"}
Score : 2986
stopped : False
Local maintenance : False
crc32 : 98aed4ec
Host timestamp : 15340475
Extra metadata (valid at timestamp):
metadata_parse_version=1
metadata_feature_version=1
timestamp=15340475 (Fri May 12 14:59:22 2017)
host-id=3
score=2986
maintenance=False
state=EngineDown
stopped=False
[root at ovirt02 ~]#
The virtual node ovirt02 has the hosted engine vm running on it
It was some months I didn't come back, but it seems it is still up... ;-)
[root at ovirt02 ~]# uptime
15:02:18 up 177 days, 13:26, 1 user, load average: 2.04, 1.46, 1.22
[root at ovirt02 ~]# free
total used free shared buff/cache
available
Mem: 12288324 6941068 3977644 595204 1369612
4340808
Swap: 5242876 2980672 2262204
[root at ovirt02 ~]#
[root at ovirt02 ~]# ps -ef|grep qemu-kvm
qemu 18982 1 8 2016 ? 14-20:33:44 /usr/libexec/qemu-kvm
-name HostedEngine -S -machine pc-i440fx-rhel7.2.0,accel=kvm,usb=off
the first node (used for deploy with hostname ovirt01 and with name inside
oVirt web admin gui of hosted_engine_1) has other 3 L2 VMs running
[root at ovirt01 ~]# ps -ef|grep qemu-kvm
qemu 125069 1 1 15:01 ? 00:00:11 /usr/libexec/qemu-kvm
-name atomic2 -S -machine pc-i440fx-rhel7.2.0,accel=kvm,usb=off
qemu 125186 1 2 15:02 ? 00:00:18 /usr/libexec/qemu-kvm
-name centos6 -S -machine pc-i440fx-rhel7.2.0,accel=kvm,usb=off
qemu 125329 1 1 15:02 ? 00:00:06 /usr/libexec/qemu-kvm
-name cirros3 -S -machine pc-i440fx-rhel7.2.0,accel=kvm,usb=off
I also tested live migration with success.
Furthermore all the 3 ESXI VMs hat are the 3 oVirt Hypervisors have still
in place a VMware snapshot, because I was making a test with the idea of
reverting after preliminary testing and this adds further load...
see here some screenshots:
ESXi with its 3 VMs that are the 3 oVirt hypervisors
https://drive.google.com/file/d/0BwoPbcrMv8mvWEtwM3otLU5uUkU/view?usp=sharing
oVirt Engine web admin portal with one L2 VM console open
https://drive.google.com/file/d/0BwoPbcrMv8mvS2I1eEREclBqSU0/view?usp=sharing
oVirt Engine web admin Hosts tab
https://drive.google.com/file/d/0BwoPbcrMv8mvWGcxV0xDUGpINlU/view?usp=sharing
oVrt Engine Gluster data domain
https://drive.google.com/file/d/0BwoPbcrMv8mvVkxMa1R2eGRfV2s/view?usp=sharing
Let me see and find the configuration settings I set up for it, because
some months have gone and I then had little time to follow it...
In the mean time, what is the version of your ESXi environment? Because
settings to put in place changed form version 5 to version 6.
What are particular settings you already configured for the ESXi VMs you
plan to use as oVirt hypervisors?
Gianluca
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20170512/c8ba291a/attachment.html>
More information about the Users
mailing list