So, we finally (started) moving that 2010 rack of servers to 3 servers

Hi everyone here,

I opened a thread some time (years) ago, obviously,without activity, it was closed after some time (ref : Moving a full rack of ~2010 gear to newer hardware ) .

tl,dr; Works like a charm, load is barely noticeable, 1/3 of the VMs moved.

2x

  • HP G8p with SSDs , controler in IT mode, boot on CF on board, 2x E5-2650, 256Gb of ram, 2x 10G on X520DA2
    1x
  • SM 1114CS-TNR, full NVME, with dedicated boot drives, 256G and also 2x 10G.
    Ceph, and a hacked backup script that does VM snapshot to a remote storage, and a recent Proxmox Cluster with live migration between hosts.

So price wise, the Epyc machine was a rough cost, but the 2 G8s were cheap, as the only expense were the SSDs and the ram extension. I think it’s interesting to notice that IO wait will cause a lot of performance issues, and that it’s hard to diagnose with older Proxmox versions.

We plan to get a second epyc SM to replace the a 1st G8, and we will plan to do the same next year. With price going down, we expect the upgrade will be simpler.

A simple pleasure migrating the VMs was the amount of time it took to backup, usually the restore took a 3rd or a 4th of the time to restore.

Ready to share more if anyone have some questions, I’ll probably have to make a thread about our backup script.

2 Likes

RBD mirroring not an option? I guess if not, you gotta be creative to backup Ceph and do it really old-school style.

could have been, but we would have needed another ceph / rbd storage ready, which we don’t have (yet). But that would simplify backups greatly, fully agreed.

1 Like