Hi Martin. Thanks for the patch. It may be relevant but I have made a significant discovery: My XML output from gmetad differs from my XML output for gmond in the case of cpu_wio. Gmond (solaris) gives me a sensible value (between 0 and 100) but gmetad (Linux) gives me some crazy number. The cpu_wio TMAX parameter also differs. Gmond gives me 90 and gmetad gives me 3800. Sounds fishy...shouldn't TMAX be consistent?? I'm running gmetad 2.5.5 on Linux Enterprise 3.0. I know that cpu_wio wasn't previously available on Linux....any significance??
By the way, how can I obtain 2.5.7 without having to deal with CVS? Maybe I should just focus on testing with 2.6.0. ---- Yemi > -----Original Message----- > From: Martin Knoblauch [mailto:[EMAIL PROTECTED] > Sent: Tuesday, June 01, 2004 2:45 AM > To: Adesanya, Adeyemi; ganglia-developers@lists.sourceforge.net > Subject: Re: [Ganglia-developers] Adding cpu_wio to universal metrics > > > --- Adeyemi Adesanya <[EMAIL PROTECTED]> wrote: > > On 5/27/04 8:03 AM, "Martin Knoblauch" <[EMAIL PROTECTED]> wrote: > > > > Hi. > > > > If cpu_wio is a percentage does anyone know why some of my Solaris > > hosts report a value far exceeding 100? (Ganglia 2.5.5). > > > > ---- > > Yemi > > > Hi Yemi, > > could you try out the patch below (against 2.5.7, but should > apply to 2.5.5). There is a pretty stupid bug which > would/could show up in 64-bit mode. Just shows that > prototypes are useful. > > Not sure that it solves your observation, because I never > saw it in 64-bit mode on my SLoaris-8 boxes. > > Martin > > diff -u solaris.c-06012004 solaris.c > --- solaris.c-06012004 2004-06-01 09:57:59.901916000 +0200 > +++ solaris.c 2004-06-01 11:36:37.129857000 +0200 > @@ -345,14 +345,7 @@ > * stuff to match. > */ > > -long percentages(cnt, out, new, old, diffs) > - > -int cnt; > -int *out; > -register long *new; > -register long *old; > -long *diffs; > - > +long percentages(int cnt, int *out, register unsigned long *new, > register unsigned long *old, long unsigned *diffs) { > register int i; > register long change; > @@ -422,8 +415,8 @@ > int cpu_states[CPUSTATES]; > unsigned int ncpus; > static unsigned long cpu_old[CPUSTATES]; > - unsigned long cpu_now[CPUSTATES]; > - unsigned int cpu_diff[CPUSTATES]; > + static unsigned long cpu_now[CPUSTATES]; > + static unsigned long cpu_diff[CPUSTATES]; > unsigned long diff_cycles = 0L; > unsigned long time_delta = 0L; > double alpha, beta; // lambda lambda lambda!!! > > > > > > ===== > ------------------------------------------------------ > Martin Knoblauch > email: k n o b i AT knobisoft DOT de > www: http://www.knobisoft.de >
