From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr; rv:1.7.2) Gecko/20040804 Description of problem: Gulmd use the default network interface to communicate with other nodes, instead of the network interface I set in nodes.ccs Version-Release number of selected component (if applicable): GFS-6.0.2.20-2 How reproducible: Always Steps to Reproduce: 1. Get a cluster with one gulm server. 2. Set nodes.ccs with the network interface I want (eth1): ip_interfaces { eth1 = "192.168.1.124" } ... 3. Start GFS. 4. Start a program that use a lot of locks/unlocks between the nodes 3. Check with tcpdump eth0 and eth1 on port 41040(gulm) that gulm use eth0 or eth1. Actual Results: tcpdump show that gulm use eth0 (192.168.0.124) instead of eth1 (192.168.1.124). Expected Results: tcpdump should show that gulm use eth1 because I set eth1 in nodes.ccs and I set the IP of eth1 in nodes.ccs Additional info: The relevant code is in: gfs-6.0.2.20/gulm/daemon/main_main.c:157:set_myID(). This piece of code have changed in the CVS version: cluster/gulm/src/main_main.c:154:set_myID(). But need to use a stable version of GFS on a Linux kernel 2.4.
you did remember to add the following line to each node entry, right? usedev = "eth1"
No, I didn't know the option "usedev". And the documentation http://www.redhat.com/docs/manuals/csgfs/admin-guide/s1-sf-nodes.html does not mention it, but I just find the documentation errata http://www.redhat.com/docs/errata/csgfs/ which mention it. Thank you for your help. I guess you can close this bug now.