If I have some time I might try, but just a warning, I don't really know
php at all, just know a good amount perl and some basic html.

I tried the updated webfrontend, It makes much more sense with the load
normalized to the number of CPUs instead of number of nodes.  I also
noticed that you committed some patches for gmetad.  Was that supposed
to fix the synching problem or are you still working on it?  I tried it
out, but didn't have much luck, it would only run for a few minutes then
some clusters would stop updating along with the top level grid.  Is it
some kind of thread locking problem?

Finally, is there a known segfault or other crashing problem with
gmond?  We have it running on a little over a thousand nodes right now
and every day about 1-2 gmond daemons die with no indication of any
problems in the system log.  I could try turning debugging on for a few
days on one of the sub-clusters and see if I can catch it.  What debug
level would be sufficient without filling up the system log too fast? 
What are the different debug levels that ganglia distinguishes anyway?

~Jason


On Tue, 2003-05-27 at 15:12, Federico Sacerdoti wrote:
> On Friday, May 23, 2003, at 01:34 PM, Jason A. Smith wrote:
> 
> > Also, I noticed that the Avg Load percentages are derived from the
> > number of hosts up.  I think it would be more accurate to use the 
> > number
> > of CPUs instead, at least if it is going to be reported as a percentage
> > instead of an absolute value.
> 
> Yes, I agree. The change will be in CVS soon.
> 
> The pie chart requires the GD library to be compiled into your mod_php 
> apache module. If you wanted to help, you could make a html tables 
> version that gives the same information, as not everyone has the GD 
> module for the pie chart.
> 
> What we need is this (made with an HTML table):
> 
> Cluster Load Percentages
> 
> 100+  |XXXXXXXXXXXX
> 75-100        |XXXXXX
> 50-75 |XXX
> 25-50 |XXXXXXX
> 0-25          |
> down  |X
> 
> There are several variations on this theme that would work (all colors 
> in one bar to illustrate their sum is 100). The "width" table cell 
> attribute would be useful here.
> 
> Let us know if you can help with the coding of this project,
> Federico
> 
> Rocks Cluster Group, San Diego Supercomputing Center, CA
-- 
/------------------------------------------------------------------\
|  Jason A. Smith                          Email:  [EMAIL PROTECTED] |
|  Atlas Computing Facility, Bldg. 510M    Phone:  (631)344-4226   |
|  Brookhaven National Lab, P.O. Box 5000  Fax:    (631)344-7616   |
|  Upton, NY 11973-5000                                            |
\------------------------------------------------------------------/


Reply via email to