Proxmox Upgrade Time

My home office lab server was running Proxmox 4.x which was clean installed ~late-2016 onto suitable spare hardware (quadcore AMD, 8gigs ram, pair of 1Tb SATA Drives) via standard 'custom manual' process (ie, Install minimal debian 8 first and manually config software raid, then install proxmox after via apt source edits). This is of course quite painless and smooth.

In the past, upgrade of proxmox 'in place' had been a bit of an adventure so I was curious how it would go on more recent flavour.

This was done and the net result is that - it is painless, even for a box running atypical config (ie, software raid) under the hood.

General steps followed were simply as per the docs at Proxmox, as per https://pve.proxmox.com/wiki/Upgrade_from_4.x_to_5.0

and it really was smooth sailing, entirely painless. Clearly it takes a little while to download all the update packages - approx 800 megs - and then a while to install everything. But it really did come down to,

  • Power off all VMs to be safe.
  • Make backups of VMs to other media if you are paranoid / worried / serious
  • start the in place upgrade, first to get up to date on current
  • then edit and move to next debian version suitably
  • then change proxmox sources as well
  • pull updated packages, install, reboot. Wait patiently while downloading and installing. Go have a coffee, watch a movie, whatever.
  • once it is done, login via SSH and then webUI and bingo, happy days.
  • Do a cleanup via apt on command line once done in case there are leftovers to be tidied up / removed.

Generally speaking, Proxmox 4 to Proxmox5 is more about the debian version change under the hood - the core cluster management toolkit is more or less the same (apparently) to the point that proxmox docs suggest you can tolerate a 'hybrid upgrade' on a cluster, ie,

  • if you have cluster of 3 x proxmox4 nodes
  • upgrade node 1 from 4>5 and now you have a hybrid cluster
  • then upgrade node2 from 4>5, still is hybrid
  • finally upgrade node3 from 4>5, now is all pure prox5 cluster, happy days.
  • but the fact this can even be done is quite nice, and makes the process more seamless (ie, lets you live-migrate VMs out of nodes in clean manner to reduce downtime for production VMs while doing the hybrid cluster upgrade. Woot. Lets see you do this on VMware without banging your head on the wall.