Online add new node to work cluster, and problem clvmd

[Melanxolik]

There was a problem with the addition of a new node to the existing cluster proxmox.
After you add a node, it rebooted the problem starts with the fact that disappears in rgmanager clustat, also a new node after the restart can not enter the clvm domain.
To solve this problem we have to restart all nodes at the same time, including the new, then everything works again like clockwork.

We use the iscsi + clvmd + ext4, the problem is with the addition of a new node, then after a while starts to work on other nodah fencing, resulting cluster inoperable.

There are some online procedure of adding a node, or is the problem clvmd?

If the issue can be considered only on a fee basis, verify that it does take us? Now for all nodes bought the license for 1 year without support.

 

[mir]

It could also be a problem with fencing. What do you use for fencing and how is it configured?

 

[Melanxolik]

All nodes use or overwrite ilo layer or module from the main supply ipmilan, problems with the fencing has not noticed.

group_tool ls

fence domain
member count 11
victim count 0
victim now 0
master nodeid 2
wait state none
members 1 2 3 4 5 6 7 8 9 10 11

dlm lockspaces
name rgmanager
id 0x5231f3eb
flags 0x00000000
change member 11 joined 1 remove 0 failed 0 seq 3,3
members 1 2 3 4 5 6 7 8 9 10 11

name clvmd
id 0x4104eefa
flags 0x00000000
change member 11 joined 1 remove 0 failed 0 seq 6,6
members 1 2 3 4 5 6 7 8 9 10 11

10 The 11th and a new node with which we have repeatedly had problems. Maybe we're not all the methods for fence agent that is used on our side of the servers from here is the behavior of the system, but we do not see that the state of la 'fence_tool ls' change during operation by adding nodes.
Can someone lead a more detailed algorithm with respect fence whether?