matt massie wrote:
steve-to get gexec working with 2.5.0... you need to compile the monitor-core with --enable-gexec. the reason gexec thinks there are no hosts up is because no gmond is multicasting that it is available.
[insert Sims-style "Uh oh, you just busted the dishwasher!" sound]That's quite a relief, because I thought that perhaps the metric numbers weren't lining up anymore. The metrics *ARE* showing up on the 2.4.0 gmonds (no, I haven't gotten them to upgrade them all yet - soon... very soon...), but they aren't showing up on the 2.5.0 muted gmond that I use as a gmetad data source. Most of the metrics line up, except "gmond_started" reports as 1 or 2 instead of the proper number... ruh roh, raggy!
The metric threshold on gexec indicates that it's rebroadcast between every 3 and 5 minutes (180 - 300), so I should have seen them by now. Something's definitely up.
I did try recompiling that gmond with --enable-gexec and with the metric.h from the 2.4.0 tarball (to make sure they'd line up), but I'm getting the same behavior. I am pretty sure that the gexec flag is being received but processed as something else.
Actually I think I noticed this back during the 2.5.0 release but only within the last week or two are people saying to me, "Hey, I hear this Ganglia thing comes with an execution environment..."
if ( host_up && gexec_available ) { use_that_host; }
Yeah, I saw this code (and oddly, it's almost exactly as you've pseudocoded it). I was hoping it was an "oops, gexec has some stuff in the XML parser hardcoded that breaks when you add TN/TMAX..." situation. Doesn't look like it.
if you run "gstat -a" you'll see that all your hosts are showing Gexec OFF.
Only from that node. They appear elsewhere (2.4.0 gstat, of course, not that it matters since gstat just parses the XML output).
brent (the author of gexec) and i talked on the phone yesterday. he is happy to lend informational help about the code but doesn't have much time to do any coding on it. i told him that i wanted to revisit gexec for ganglia 3 and modernize it to take advantage of the new features.
Yay modernization!
if you want to jump into it go right ahead. btw, i loved the earlier message about the ganglia "community" idea.
Gotta love those 9am pre-wading-into-work emails.
i'm in the final stretch for building the internal data structure for ganglia 3. i'm also preparing a presentation for the linux users group at lbl (http://lug.lbl.gov)... they asked me to speak for 1.5 hour. you don't mind if i quote you on the ganglia community model .. do you? :)
90 minutes. Whew. I hope you're bringing visual aids. Lots of them. Quote away, you'll need something to fill that time. :P Good luck...
