As described on the
Ganglia wiki page (or
IBM's related post), there are currently two ways to add customized metrics to ganglia (3.1.7):
- Spoofing with gmetric
- Writing loadable metric module with Python or C
Every time when we modify the module (or simply don't need it anymore), the old module will stay in the system, and a new figure corresponding to the module will be added (not updated) to the Ganglia front end. To remove the unused metrics, two approaches could be used here:
- For gmetric: if you go through the man page of gmetric, you probably would notice the -d (--dmax) option which sets the expiration time of a metric if no new message is received within dmax seconds. It's default value is 0 which means this metric will never expire. To remove the unused metrics, just simply re-run the gmetric command with --dmax set to a positive integer.
- For loadable modules: there is no simple way to do the job. Since each gmond multicast (default way) its information to the collector (gmetad), it probably has to kill not only the gmetad process and gmond process where this metric originates, but all gmond processes related to this multicast. The safest way is to kill all gmond and gmetad processes (if it's possible and easy to do). See this post for more details:
No comments:
Post a Comment