2 nodes HA migration

[Alessandro 123]

I've created a test cluster with 2 nodes.
Added one pool (pool0) with both nodes set
Added one resource, poiting to that pool, with vm 100 set
Fencing is the software watchdog.

If I power down a server via SysRq (echo o > /proc/sysrq-trigger), proxmox detect the missing node (the web interface marks that with the red icon) but VM is not started on the other host.

These are the 3 latest lines from the log:

Sep 27 12:42:07 pve pve-ha-crm[2645]: status change master => lost_manager_lock
Sep 27 12:42:07 pve pve-ha-crm[2645]: watchdog closed (disabled)
Sep 27 12:42:07 pve pve-ha-crm[2645]: status change lost_manager_lock => wait_for_quorum

Ok, I need at least 3 nodes, but is possible to skip the quorum check as I'm just doing some testing....
I don't have a third node available.

 

[tom]

you need three nodes for HA.

your test shows you why.

 

[Alessandro 123]

Ok, but I'm just testing. I don't have a third node here.
Isn't possible to hack the corosync config to allow a two node cluster ? It's just a test, i'll never put this in production

 

[Alessandro 123]

I've solved by adding "two_nodes:1" to /etc/pve/corosync/corosync.conf inside "quorum" node

HA migration is working properly.

 

[tom]

This is not a solution, its a way to hell.

Please use at least there nodes.

 

[Alessandro 123]

This is not a solution, its a way to hell.

Please use at least there nodes.

I know, but as I wrote: it's just a test.
I don't have anything important on these servers and these server will be decommissioned tomorrow.

I'll never put this in production, i'll use different servers on different networks (10GbE)