[ovirt-users] Re: How do I share a disk across multiple VMs?

This turned into quite a discussion. LOL. A lot of interesting points. Thomas said --> I think Sandro already spoke into this a little bit, but I would echo what they (he? she?) said. oVirt is an open source project, so there's really nothing preventing any of us from jumping in and assisting where we can. Granted, I'm not much of a software developer, but eventually, I could see how I can contribute my time in some ways. Replying to emails on the mailing list, providing engineering input on system level decisions, testing RC releases, fixing / debugging Ansible scripts, etc... (I love ansible!), helping to update documentation, etc... Sandro Said --> My hosts are running RHEL 8.3, and I have no plans to move to something different. Ironically, though, the vast majority of my VMs are running Ubuntu. Off topic, but something to address: We need a stable ovirt-guest-agent package. This doesn't seem to be working for me, although I'll take a look at it more closely again when I have some time: https://launchpad.net/ubuntu/focal/+source/ovirt-guest-agent Thomas said --> Serious question: What's preventing you (or anyone) from just spinning up new storage with NFS, SCSI or whatever, mounting to the engine, and migrating your VMs to that new storage? Correct me if I'm wrong, but HCI is more of a philosophy and a framework than anything. There's nothing that prevents us from moving away from an HCI model. In fact, I'm getting ready to do something that will already start to move my environment in that direction. I have a 4th server that I originally bought last fall for testing & spare parts, but I've decided I want to put it into the datacenter, and use it as a backup destination, as well as a second DNS server for the overall environment. I have 3 spinning disks arriving next week that I'll put into a RAID 5 on that server, and my plan is to then build two different NFS mount points, and expose those NFS shares to the oVirt Cluster. I'm also half tempted to add that server as a host to the cluster as well, because it has 40 cores in it that would otherwise just sit there unused. I'll use 1 NFS share as a Backup domain, and I'll use another NFS share to store images that don't need ssd speeds - such as ISOs, and such. Sorry for the rabbit trail, but this plan I think is a classic example of my point that HCI is a philosophy, not a hard and fast rule. You can do whatever you want. If you don't like HCI, why can't you move to something else? Gianluca said --> That's interesting to hear. Out of curiosity, why keep gluster around? And also, why hasn't there been much of an effort to support Ceph in a HCI environment? I actually had someone who I know is a Red Hat TAM advise me (off the record, off the clock, not through official Red Hat channels) to stay away from Gluster as well. After my research & testing, though, it was the *only* way I could see how I could deploy my environment on the limited budget that I had (and still have). Eventually, yes, I would like to get better storage. One idea that comes to mind that would be stable and still relatively cheap is to grab a couple of Synology NAS devices, stick ssds into them, and put them into an HA pair, and expose that storage as an iscsi mount. - David Sent with ProtonMail Secure Email.