Description of problem: I wanted to use a simple fence method with no arguments (like /bin/true). Part of the cluster.conf file looks like this: <clusternode name="link-10" votes="1"> <fence> <method name="abc"> <device name="123"/> </method> </fence> </clusternode> ... <fencedevices> <fencedevice name="123" agent="logger DIE, FOO"/> </fencedevices> The problem line is: <device name="123"/> It can be unproblemmed by adding something, like: <device name="123" myargs="getsome"/> The problem is that when a fence operation is attempted ccsd and fence_tool join together to consume 100% CPU and hang: 2767 root 16 0 14536 1356 1136 S 60.6 0.3 6:53.13 ccsd 2775 root 25 0 1648 308 252 R 39.3 0.1 4:36.51 fence_tool Version-Release number of selected component (if applicable): 6.1 RPMS built Wed 15 Dec 2004 01:13:08 PM CST How reproducible: Yes. Steps to Reproduce: 1. Configure a fence device with only a "name" key/value 2. Start fence_tool -D 3. Actual results: Hangs on the first attempted fence operation, which is: "fence_tool: unfence ourself" Expected results: Don't do that. Additional info:
note to dct: ccs_get_list is now available, which will not barf in the presence of the above situation.
fenced now uses ccs_get_list to avoid the problem
Verified against 2/28/2005 build.