Corosync crashed if node has more then one device in the same subnet.

This document (7008804) is provided subject to the disclaimer at the end of this document.

Environment

SUSE Linux Enterprise High Availability Extension 11 SP1

Situation

Using broadcast or multicast in corosync if more then one device linked to the same subnet, corosyn sometimes causes a crash.
The following example shows a configuration where this can happen:

# ip a | grep eth

2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000
    link/ether 00:16:3e:0f:4c:2d brd ff:ff:ff:ff:ff:ff
    inet 192.168.42.20/24 brd 192.168.42.255 scope global eth0
3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000
    link/ether 00:16:3e:0f:4c:78 brd ff:ff:ff:ff:ff:ff
    inet 192.168.42.120/24 brd 192.168.42.255 scope global eth1

# grep -A 4 interface /etc/corosync/corosync.conf

        interface {
                bindnetaddr:    192.168.42.0
                mcastaddr:      226.94.1.2
                mcastport:      5406
                ringnumber:     0

Resolution

To avoid problems by having more then one devices in the same subnet, it is possible to use udpu for the network protocol. Here it is possible to use "unique" ip adress for each node (see man corosync.conf for details).

Additional Information

Corosync use the bind network address (bindnetaddr) to find out which network device should be used for his communication. If there are more then one device linked to the same subnet, corosync will be use a device by chance. But there is no guarantee that this device is always the same. Also it can't switch to the other device if the working one will left.

Disclaimer

This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.