Open stephensje opened 8 years ago
This is the default behaviour (i.e. loading the module means enabling the metrics too).
In some debian packaging (.deb), they avoided this by renaming the configuration file:
Note:
- mod_multicpu is not enabled by default, you must rename the config:
cd /etc/ganglia/conf.d && mv mod_multicpu.conf-sample mod_multicpu.conf
I don't think that this was from the ganglia-gmond-python package, though ?
http://pkgs.fedoraproject.org/cgit/rpms/ganglia.git/tree/ganglia.spec?h=el6
The mod_multicpu seems to be included in the main ganglia-gmond rpm:
cp -p gmond/modules/conf.d/*.conf $RPM_BUILD_ROOT%{_sysconfdir}/ganglia/conf.d/
rm -f $RPM_BUILD_ROOT%{_sysconfdir}/ganglia/conf.d/{modgstatus,example}.conf
Maybe ask the Fedora EPEL maintainer of ganglia, to make it opt-in ? (and there does seem to be a similar issue, with the 9 *.pyconf files)
I administer a Linux cluster running CentOS 6.6 and recently installed Ganglia 3.7.2. The default gmond setup is working as expected. A few of the nodes have GPUs, so I installed gmond-python so I could later add the nVidia metrics. The installation went fine, but I noticed that ganglia-web now shows all the multicpu metrics, even though I did not touch the multicpu.conf file (where these metrics are all commented out by default). I can only make them go away by renaming the multicpu.conf file.