@yehor@mastodon.glitchy.social
2026-07-17 06:41 UTC
Replies (3)
-
@monotux@hachyderm.io 2026-07-23 08:20
@yehor@mastodon.glitchy.social https://blog.hofstede.it/hetzner-proxmox-cluster/ @Larvitz@burningboard.net did something similar
-
@iMeddles@mastodon.eddmil.es 2026-07-17 06:56
@yehor@mastodon.glitchy.social don't do it. Proxmox uses corosync for cluster communication, and it's incredibly sensitive to network latency and dropped packets, unless your network link to the hosting provider is superb, you're going to end up with the server getting fenced constantly. A better idea is to run it as a stand alone sever, and use either Proxmox Datacenter Manager, or Pegaprox, both of which will let you manage all the servers through a single interface, and allow live migrations between them, without the servers having to be clustered together.
-
@yolo@defcon.social 2026-07-17 09:02
@yehor Yo probe exactamente eso: un nodo Hetzner CX22 con Proxmox unido al cluster casero via WireGuard. Lo que aprendi: (1) La latencia importa — si el VPS tiene >20ms de ping, HA se vuelve problematico porque el quorum puede fallar. (2) Usa corosync + votequorum con expected_votes: 2 y un tercer nodo witness ligero (RPi) en casa. (3) Migracion en vivo no funciona bien por latencia — usa backup/restore. Para servicios no criticos funciona perfecto.