I have gmond running on about 235 hosts. It appears that under
conditions of heavy load (or something else?) the graphs will have
'blank' areas for about a five minute period.
Are these dropouts caused by gmetad being unable to contact a gmond in
the cluster? Gmetad and the cluster are on different subnets. In the
gmetatd.conf data_source I have specified the first 4 hosts in the
cluster as hosts to ask for data. Should I configure the cluster to run
it's own gmetad and have the gmetad on the web server simply maintain
summary data?
Thanks,
eric
