[PVE-User] High Availaility / Fencing in Proxmox VE 4.x

[Uwe Sauter]

Dear all,

I'm a bit confused by the Wiki pages regarding high availability [1] & [2].
I would appreciate if my questions could be answered. (Searched the mailing list
archive but didn't find threads regarding HA that are new enough to cover v4.x)

* When does a Proxmox cluster become a HA cluster?
  In v4.x do I still have to configure HA as described in [3] or is it HA-capable
  once a cluster with at least three nodes has been formed and all I have to do is
  to add resources (VMs) to be managed by 'ha-manager'?

* How does fencing work in Proxmox (technically)?
  Due to fencing being based on watchdogs I assume that some piece of software
  regularly resets the watchdog's clock so that the node isn't rebooted automatically
  by the watchdog. Is this correct? Which software is responsible?

* What criteria are used in the decision making whether a node needs to be fenced?
  I'm asking this as I currently have a 4-node cluster with Ceph as test bed and
  fencing doesn't work the way I expected. I'm using a separate interface for
  Ceph traffic. If I disable that particular link I loose communication with the
  Ceph cluster and therefore the block device for all running VMs on that host.
  I'd expect that this would result in fencing this particular host but that isn't
  the case. My setup might be missing correct configuration for this to work but
  I didn't find documentation on how to define fencing criteria.

* Does having a second ring for corosync ("ring1") do anything to do with HA?

* Do I assume correctly that fencing using the IPMI protocol in combination with
  the server's BMC is not supported in Proxmox 4.x?

* At the end of [1] it is recommended to do testing before going into production.
  What kind of tests are recommended? How to simulate network errors / disk failures /
  what else needs to be tested…?


Looking forward to getting your answers.

Regards,

    Uwe

[1] https://pve.proxmox.com/wiki/High_Availability_Cluster_4.x
[2] https://pve.proxmox.com/wiki/High_Availability
[3] https://pve.proxmox.com/wiki/Fencing