From: Steven W. <sw...@il...> - 2002-06-26 00:53:15
|
Ryan Sweet wrote: > On Tue, 25 Jun 2002, Steven Wagner wrote: > > >> ... I just got gmond-current to run on Tru64!!! >> >>Sure, I was using a hacked osf.c based on the horrifyingly out-of-date >>cygwin.c ... and sure, the netcode doesn't seem to bind to a network > > > which leads me to ask, what might be necessary to get cygwin to build and > run gmond? Since Cygwin doesn't have multicast support is it desirable > that gmond could accept an arg to tell it to use a non-multicast mode > (instead use tcp with data collection/reporting to a specified set of > hosts)? Or does it make more sense to try and port gmond to win32? I > admit, I haven't looked at the gmond code closely enough to even come > close to answering this myself, but I have done similar exercises with > other sgi/solaris programs in the past. > gmond doesn't seem to (by design) have any unicast metric gathering capabilities. It should be possible to report local metrics to a client (i.e. you could specify it as a gmetad target), though. Anything more sophisticated than that would have to be Something Weird[tm], compared to the rest of the "family." When I first started following the Ganglia project I couldn't figure out what the piece was that "bridged" clusters, and gmetad has been put forth as that piece. So it does seem like the logical target. Otherwise it'd be time for a Win32 port. But that's one port that I ain't touching with a 5,281-foot pole. Although at my last job I know people who would have *KILLED* for an open-source win32 Ganglia-like monitoring tool. Up to and including gexec. That's one platform that I could see a *BIG* audience for... |