Hello,
This is for packetfence-9.0.1-1.el7.noarch.
We have the following lines in "var/conf/mariadb.conf" (amongst others):
bind-address=192.168.0.37
wsrep_cluster_name=pfcluster
wsrep_cluster_address="gcomm://192.168.0.28,192.168.0.37"
wsrep_node_name=net-pf1
wsrep_node_address="192.168.0.37"
(We will be adding a third server eventually.)
This is presumably generated from our "conf/cluster.conf":
[net-pf1]
management_ip=192.168.0.37
[net-pf1 interface eth0]
ip=192.168.0.37
[net-pf2]
management_ip=192.168.0.28
[net-pf2 interface eth0]
ip=192.168.0.28
However, because we are using OpenStack (OS) and the way its networking is
architected, our instances/VMs have internal-to-OS addresses (192.168/24)
that are actually live on the ethX interfaces, as well as external-to-OS
addresses (10/8) that are handled by OS 'ports' and tunnelled into OS.
Is there a way that I can have the "wsrep_cluster_address" and
"wsrep_node_address" entries be a 10.x address, while the "bind-address"
is the 192.168.x address?
I'd like the 'wsrep' items to have 10.x because some of our HA stuff will
be in a different part of the network, and won't be able to use the
internal-to-OS 192.x addresses.
I would think that the 'interface eth0' stanza would control the
"bind-address" value, but the "management_ip" would be for what other
nodes talk to. What is the difference between "ip" and "management_ip"?
--
DM
_______________________________________________
PacketFence-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/packetfence-users
