The nodes to upgrade [1]:
- [x] uffizi
- [x] pompidou
- [x] hypervisor3
- [ ] beaubourg
- [ ] branly
# Plan
Looping on each hypervisor in the following order, we will mostly move or stop vms
elsewhere then revert the state after each upgrade to bullseye:
- (for all) Pre-check:
network interface configuration to avoid reboot caveats (no network at reboot)
- 1. uffizi: sandbox hypervisor
- vms to stop: all
- 2. pompidou: it's staging infrastructure
- vms to move:
- staging workers
- workers[13-16]
-> vms moved without stopping just restarted once moved to uffizi
-> so in the end, stopped the vm properly prior to migrate to uffizi
- 3. hypervisor3:
- stopping vms:
- webapp1
- louvre
- pushkin (secondary firewall)
- workers[01-08]
- moving some vms to branly (same archi amd):
- logstash0
- bardo (to ensure front-end service for hedgedoc)
- tate
- 4. beaubourg: only somerset (secondary db) is running there and the archive webapp is
currently using belvedere db (primary) so it can be stopped without impacting the
archive)
- 5. branly:
- moving vms to hypervisor3 (same archi amd):
- pergamon
- moma
- saatchi
- thyssen
- counters1
- search1
- kelvingrove (keycloak)
- stopping and moving to another hypervisor:
- riverside (sentry)
- rp1 (rp for hedgedoc)
- stopping other vms:
- glyptotek: primary firewall (switch it as secondary first during the
duration of the upgrade)
- workers[09-12]
- getty
- kibana0
- bojimans (netbox)
- jenkins-debian1
[1] https://inventory.internal.softwareheritage.org/dcim/devices/?cluster_id=4
[2] Highlighted doc from irc discussion:
- [ ] https://pve.proxmox.com/wiki/Upgrade_from_6.x_to_7.0
- [x] T3787: Prerequisite to the upgrade to debian 11 [3]
[3] https://pve.proxmox.com/wiki/Ceph_Nautilus_to_Octopus