About a year ago I wrote a couple of posts about setting up OpenNMS with WMI (part 1, part 2), but until recently we were unable to leave this functionality enabled all the time because of a socket file descriptor leak (OpenNMS bug 3249). If we enabled WMI more and more socket descriptors would be allocated and not freed, and eventually everything would grind to a halt and monitoring would fail. So we left WMI disabled, and tested again each time we updated OpenNMS.
Happily for us, the OpenNMS folks have fixed this issue. It looks like it was fixed in OpenNMS 1.8.3, and I confirmed it fixed in 1.8.4, the current version. So I'll be writing additional posts about implementing weird and wonderful WMI counters as we get them setup -- likely the first set will be for Microsoft AppFabric Caching, which we're now testing.