From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.8) Gecko/20050610 Firefox/1.0.4 (Debian package 1.0.4-3) Description of problem: We are using IBM blades to host a small cluster. The blades come with Broadcom's 5700 series network interfaces (can't remember the exact model). We can use either Broadcom's own bonding + VLAN drivers, or the ones that ship with the linux kernel. The Broadcom VLAN+bonding driveres experience a stability issue so we have to use the Linux standard drivers instead. However, using the linux drivers result in the NICs being named to bond0.4010, that is, bond0. + a VLAN ID appended to the bond0. Problem: RHCS uses ifconfig to list the NICs, and ifconfig only doesn't show the full NIC name if it is too long. Thus the error message "Error: IP Address not on same network as this host". Another problem with RHCS is that the regex matching the interface names from ifconfig output doesn't match a dot in the NIC name... Version-Release number of selected component (if applicable): redhat-config-cluster-1.0.2-2.0 How reproducible: Always Steps to Reproduce: 1. Use the bonding.o and 8021q.o modules on RHCS cluster members 2. Try to use one of the bond0.xxxx interfaces for the RHCS traffic, xxxx being a four-digit VLAN id 3. Actual Results: The mentioned error message Expected Results: The cluster should properly have had the nodes set up Additional info:
I *think* the option to use netlink as the service IP internals would fix this corner case as far as the back-end is concerned.
At this point, this change will not be made to RHEL 3 clustering.