Ovirt newbie here - installing v 4.4.4 (well, trying to!)
Following my post about failing Gluster setup, I can't get the self hosted engine to
deploy. I'm installing on my HP DL380p G6. I have 2 disk 170GB Raid 0 for OS and 6 x
330GB disk Raid 5 for Gluster. DNS all set up, but Im realizing I'm missing something
here. I have a feeling my problems with Gluster/Engine were caused by incorrect network
setup. Seems I'm not the 1st to fall at this hurdle - some people in the community
are saying RH have made it's deliberately difficult to get these 'free'
set-ups working. It does feel a little like that!
Most instructions just say that a 'self hosted, hyperconverged, single node setup
requires 2 nics' - and that's about it! That is the 'Networking
Pre-requisites'!
I've had some help on here and Reddit which eventually made me find a solution some
had had to use - editing lvm.conf to make sure my drive (sdb) was being blacklisted
correctly.
So, with that, Gluster installs, with following DNS settings:
The kvm host (
ovirt-kvm.whichelo.com) is fixed ip 192.168.0.40 on my 1st nic (Enp0s7)
ovirt-engine.whichelo.com - 192.168.0.50
and
ovirt-gluster.whichelo.com on 192.168.0.60 - I created I VLAN linked to the nic i want
to use for gluster (Enp0s8) and gluster install worked.
So now hosted engine won't install,, and pretty sure it's because I don't know
how to set the network up properly. I'm seeing virbr0 coming up with different
ip's - sometimes 192.168.1.1, sometimes 192.168.222.1. From what I've read, this
is something to do with the Engine's network, but I really don't know!
Am I still missing something? I can't find any decent instructions on how to do this -
how (exactly) to configure the 2 'minimum required' nics?
I came to Ovirt after realizing Oracle were dumping their own Virtualization platform in
favour of KVM. Couldn't set it up from Oracle so moved to Red Hat - that was a no,
and I was at the point where it made most sense to just run an Ovirt node for my KVM.
I've got much further with Ovirt, but instructions do not work as easily as they
looked! (example 'just click on single node hyperconverged' which didn't work
straight off the bat until removing that LVM filter, which took days to find out!
The single node HC server is a great match for a home lab/server like mine. I don't
want another server, let alone another 2! I'm doing this partly as hobby, but also to
update my skills during lockdown. Surely if we can get people using things like this at
home, they're more likely to end up using it for work one day?
Any help, or just pointing in the right direction, gratefully received! Hopefully I'm
a biit clearer here. Happy to provide any logs or anything else....
Thank you!