Loss of Quorum

PorxDewy

New Member
Mar 5, 2015
11
0
1
I have two servers in my test lab running proxmox 3.4. I have configured them as a two node cluster:
1: Dell PowerEdge 860 && 1: HP Compac dc7800 Sff pc. Both 64 bit and both support VT and VT is enabled in bios.
All updates and upgrades done. All host file entries done and network/interfaces configured for my needs.
Also installed are openvswitch-switch openvswitch-common.
This is not a problem.
pve and quorum - Copy.PNG

Everything is ok for EXACTLY 7 minutes after boot from screen below(just a downloaded example not my actual screen) and yes I have timer this with precision.
Blue Proxmox screen.png

Then I get this (edited by me )

quorum - Copy (2).PNG
This happens exactly 7 minutes after boot every time, regardless of how many times I reboot.
I have a feeling this is a multicast problem but have no idea what exactly.
I have installed ssmping and on running this get
multicast *********************
unicast ********************
multicast *********************
unicast ********************
multicast *********************
unicast ********************
multicast *********************
unicast ********************
multicast *********************
unicast ********************
so that is working.

I have tried re-installing proxmox and still the same .
I would be glad of any solution to the problem.
Regards.
 

Attachments

  • quorum - Copy.PNG
    quorum - Copy.PNG
    154.4 KB · Views: 11
  • quorum.PNG
    quorum.PNG
    157.7 KB · Views: 13
Last edited:
Hello PorxDevy,

In order to be sure that multicast is functioning correctly try also to have a look at the packages sent/received in both nodes by

Code:
tcpdump -e -n -i eth0 | grep "239\.192\.181\.106"

Adapt in the above interface name and multicast IP if needed!

Have a look if all packages sent from one node arrive at the other!

Also a look into syslog (/var/log/syslog) may help!

Kind regards

Mr.Holmes
 
With 2 nodes, configure unicast in the cluster is more efficient in terms of networking performance, and also avoid that this problem continue.

If you have a switch managed, also you should configure the switch for that the multicast packages can be transmitted between hosts. But with unicast configured in your PVE cluster, this isn't necessary.