From: Federico S. <fd...@sd...> - 2003-05-20 04:01:13
|
Jason, These are all good comments. The grid summary stuff is an issue that I'm working on now and I'm pretty sure I know what it is. The small graphs are too small, I agree. I was playing around with solutions to the "how to I visualize a big cluster" problem, but they need to be bigger. Thanks for the comments, I plan to do some more work on the code late this week. -fds On Monday, May 19, 2003, at 08:36 AM, Jason A. Smith wrote: > I just tried the latest gmetad & webfrontend from cvs to test out the > improved page load speed and other features and have a few comments: > > First, the grid summary totals appear to be messed up in gmetad. Both > the rrds and xml data constantly display/report fluctuating numbers. > Is > this a known bug, have you seen this in your tests? > > I like the new more compact cluster view that gets rid of the colored > node icons and just colors the plots instead, but I think the small > graphs are a little too small. They are so small that often rrdtool > doesn't even add any labels to the horizontal & vertical axis, and the > actual plot area takes up a very small portion of the total image area. > Also, I think removing the legend is a mistake, it was useful for us to > be able to see the value "now". Sometimes one node goes out of control > reporting loads over 100 and setting the vertical range to accommodate > the highest max value means the other plots are useless, especially > with > no "now" value displayed. > > I think dead nodes should be shown at the top of the cluster list, not > the bottom, because it is important info, just like high load is. If > you have a large cluster then you don't want to have to scroll all the > way to the bottom to see if there are any dead nodes. For dead nodes, > it would also be nice if the last reported time was shown without > having > to go to the node view, and shown as a human readable timestamp. The > last heartbeat doesn't seem to work when the node is listed as dead, I > thought it worked before. Also, the text table showing the cpu and > load > averages is often too narrow making it hard to read when it wraps > repeatedly. This seems to vary depending on the browser use, I tried > the old netscape and mozilla. > > What is required for the pie charts? They don't work for me at all. > > One last minor comment, why was the default_range changed to a day? I > think an hour is much more useful and don't think the page needs to > reload itself every 300 seconds if the default range is set to 24 > hours. > > ~Jason > > PS. What is wrong with sourceforge lately, I keep getting broken > connections when trying to access their cvs server? > > > -- > /------------------------------------------------------------------\ > | Jason A. Smith Email: sm...@bn... | > | Atlas Computing Facility, Bldg. 510M Phone: (631)344-4226 | > | Brookhaven National Lab, P.O. Box 5000 Fax: (631)344-7616 | > | Upton, NY 11973-5000 | > \------------------------------------------------------------------/ > > > > ------------------------------------------------------- > This SF.net email is sponsored by: If flattening out C++ or Java > code to make your application fit in a relational database is painful, > don't do it! Check out ObjectStore. Now part of Progress Software. > http://www.objectstore.net/sourceforge > _______________________________________________ > Ganglia-developers mailing list > Gan...@li... > https://lists.sourceforge.net/lists/listinfo/ganglia-developers > Federico Rocks Cluster Group, San Diego Supercomputing Center, CA |