How to exclude a monitoring server from ganglion reports?

Question

I run ganglia in EC2 and reporting works well. I run gmetad to monitor a database cluster from an administrator reporting instance, which is a centralized dashboard for all of our systems. I do not want this instance to be included in monitoring. In admin gmond.conf, I installed:

mute = yes 

But this only makes the web interface show the host as dead. In admin gmetad.conf, I have a data_source for:

 data_source "cluster" ec2-XXXX.compute-1.amazonaws.com 

(with ec2-XXXX.compute-1.amazonaws.com is the host name of the ec2 administrator instance)

I thought setting mute=yes would remove gstat from reports, but both of them still show that the admin host (localhost) is dead.

Is there any way to do this?

Answer 1

You need to set the host_dmax attribute host_dmax value other than 0 in your gmond.conf file.

For me, an inactive Ganglia node appears first (after a restart), but then disappears after I set host_dmax for host_dmax . It is not clear why the node appears first after a restart, even if the mute parameter is set to yes.

The cleanup_threshold attribute cleanup_threshold also affect the time it takes for a host to disappear.

This is described in the Ganglia wiki :

The host_dmax value is an integer with units in seconds. When set to 0 (0), gmond will never remove the host from its list, even when the remote host has stopped reporting. If the host_dmax parameter host_dmax set to a positive number, then gmond will start the host after it does not hear from it in host_dmax seconds. By the way, dmax means "remove maximum".

cleanup_threshold is the minimum amount of time before gmond cleans up all hosts or metrics where tn> dmax aka data has expired.