I have an old VMware Essentials cluster running on 3xSupermicro SYS-1028R-WMRT servers with 2xE5-2630v3 CPUs and 256GB RAM in each. We have 30-40 VMs that perform various functions, nothing too crazy compute or IOPS wise. They’re attached to a Synology FS2017 all-flash array via 10GbE iSCSI (multipathed using two separate air-gapped 10GbE switches for redundancy. Yes, way overkill). Nodes are also dual-10GbE linked to our network on a LAG with multiple VLANs trunked.
That system has served us very well for several years but it’s time to migrate away from it over to XCP-NG on “new” hardware. New hardware mostly just because it isn’t practical to do it in place on my existing setup, which is still in fantastic shape.
My idea is to purchase 3x “new” servers: 2 XCP-NG nodes, the third used as a dedicated TrueNAS NFS appliance temporarily.
I’d migrate the VMs from VMware to the two XCP-NG nodes (the third TrueNAS NFS machine being the SR). Once that’s humming I’d reformat the Synology and deploy as NFS, moving the SR there after. Then I’d recommission the third server as a third XCP-NG node.
I do not have a huge budget for this and I’m looking at used / refurb servers. It seems like the Dell Poweredge R640 systems are the best bang for the buck out there at the moment. They’re solidly built, plentiful and affordable for spares, have great management features etc. They are several years old of course, but is there any pressing reason why I shouldn’t consider them for the migration? I don’t think the older PCIe gen 3 is an issue, we’re not doing anything higher than 10GbE on any interface. I know they’re physically longer than my Supermicros but that’s not an issue for our rack.
I guess I’m asking how many more years I can expect XCP-NG to support this hardware / modern operating systems as guests? When was the last time XCP-NG killed support for a major swath of servers (if ever) and how old was that generation?