[ovirt-users] Re: Actual oVirt install questions

On Sat, Aug 22, 2020 at 4:01 PM David White via Users <users(a)ovirt.org&gt; wrote: Let me clarify (again?): All-in-one, meaning - both (standalone) engine and actual VM processes (vdsm/libvirtd/etc. and your VM's qemu processes) on the same machine - is not supported since 4.0 and is not tested routinely and systematically. The fact that it does usually work, and the fact that people sometimes report successes with this, is very nice to know, but does not change the situation :-) If all you have is a single physical machine that you want to use for oVirt, you have, basically, two options: 1. Use HCI (meaning, hosted-engine + gluster). That's, I guess, what most people do, at least in production. When it works (usually, and we definitely aim for always!), it's the easiest way to install and get things going, but is more complex under the hood and harder to understand/debug if you do not know quite well how things work internally. 2. Use some other virtualization method - I think most people use plain virt-manager - to create separate VMs, and imitate multiple-machines setup using these VMs. E.g. one machine for the (standalone) engine, another for storage (e.g. nfs/iscsi/gluster), a few more as hosts (with nested-kvm), etc. Latter, (2.), is what we use in CI - see projects lago and ovirt-system-tests if interested. Re the imageio upload issue: If it's indeed a result of using all-in-one, then please wait until Michael's bug is fixed, or see the other thread, or see above for using not-all-in-one :-) (But feel free to keep debugging this with Michael's help if both of you feel like it!). Best regards, -- Didi