Cluster: Stonith fence device fence_vbox does not work as expected

[PaulHartwig]

Hi all,
i'm not sure if this is the right forum/thread, but the situation is as follows:

I set up a Cluster with two virtual machines (Linux Debian 11.3). The host names are drbd1 and drbd2.
The cluster setup worked fine with corosync, pacemaker and drbd to have a mirrored storage device.
Virtualbox version is: version: 6.1.32 r149290 ((0xA2CDe001U))

For stonith fence device i used the obvious fence device "fence_vbox".

command:
pcs -f stonith_cfg stonith create virtualbox fence_vbox ipaddr=192.168.0.100 username=drbd passwd=xxxxxx host_os=linux pcmk_host_map="node1:66050477-c06c-4c45-a395-eed175443541 node2:28d2c7ce-0411-48eb-9bf4-927d8639c5f6"

The host is 192.169.0.100 and it hosts different virtual machines, which are all started headless, since the host is running 24/7.

Everything is accepted by the cluster manager, the user drbd is in the vboxusers group and can log into the host and use VBoxManage.

I want that in case of a failover, one node reboots.

When a failover occurs, everything works and the login occurs, but when i look at the output of "pcs status" i see:

Command: pcs status
root@drbd2:~# pcs status
Cluster name: mycluster
Cluster Summary:
* Stack: corosync
* Current DC: drbd2 (version 2.0.5-ba59be7122) - partition with quorum
* Last updated: Sun Apr 17 12:30:42 2022
* Last change: Sun Apr 17 10:21:29 2022 by root via cibadmin on drbd1
* 2 nodes configured
* 2 resource instances configured

Node List:
* Online: [ drbd1 drbd2 ]

Full List of Resources:
* virtualbox (stonith:fence_vbox): Started drbd2
* ClusterIP (ocf:IPaddr2): Started drbd1

Failed Fencing Actions:
* turning off of drbd2 failed: delegate=, client=stonith_admin.120198, origin=drbd1, last-failed='2022-04-17 10:33:16 +02:00'
* reboot of drbd2 failed: delegate=, client=stonith_admin.60230, origin=drbd1, last-failed='2022-04-17 01:08:26 +02:00'

Daemon Status:
corosync: active/enabled
pacemaker: active/enabled
pcsd: active/enabled
-------------------------

I can login from any of the two cluster members intoo th host and do a "VBoxManage controlvm drbd2 reset"
I also tried the root user, but fence_vbox seems to do nothing but without errors or warnings.

Does anyone has an idea what i do wrong?

Thanks in advance and best regards

[scottgus1]

'stonith' and 'fence-vbox' are terms not common on the forum, or at all in the manual. Had to do a web-search, and surprisingly to me this has come up, once:

viewtopic.php?f=6&t=90003

Apparently 'stonith' is some kind of fail-over setup. Virtualbox doesn't have anything to make fail-over arrangements, except for "teleporting" https://www.virtualbox.org/manual/ch07.html#teleporting This is a completely in-Virtualbox setup which does not involve 3rd-party apps. (I have never tried Virtualbox teleporting, btw.)

I'd suggest, if you're set on 'stonith', to ask the Stonith folks.

FWIW, a Virtualbox VM could no doubt participate as part of a cluster of fail-over computers, if the VM were seen as one of the computers running the fail-over service, rather than the service that needs to be fail-over. For example, one wants to have a fail-over web page. One could run the web page in multiple computers, one or more of which could be VMs (preferably on different hosts) all drawing from fail-overed central NAS's. Running the VM as the fail-overed service could be harder, since it's a whole OS, not just an app on the OS.

[PaulHartwig]

Hi,
i'm just playing around and this is in no way anything productive.
Thanks for the answer though, i just wanted to give it a try!

[fth0]

FWIW, fence_vbox is probably not part of Oracle VirtualBox, although the man page fence_vbox(8) suggests otherwise.