Occasionally a Linux compute node will get whacky
and its automounter will cause the system load to
gradually climb to about 1000! The node is mainly
functional, so it continues to report data to ganglia.
Sometimes this happens overnight or over the
weekend before being fixed with a reboot.
The problem is that this funky load of 1000 screws
up the ganglia cluster and grid data by drastically
altering the scaling. For example, the graph for
a 100-cpu cluster will show a load over 1000, with
the true data squished into obscurity at the bottom
of the scale.
What would be nice is some script that can operate
on the rrdtool databases and "zero out" a given node
for a specified time period. Plus it would have to
correct all the aggregated data, too. Does anyone
have a solution to this situation?
--
John Saalwaechter <[EMAIL PROTECTED]>
____________________________________________________
Yahoo! Sports
Rekindle the Rivalries. Sign up for Fantasy Football
http://football.fantasysports.yahoo.com
