Auto Agregates for clusters

[appodictic]

With the work I am doing with hadoop
http://www.jointhegrid.com/hadoop-cacti ... ng_job.jsp

There are two things that ganglia does much better for large clusters.
1) ganglia is push based no configuration at all per host.
I have delt with this by writing a script that can add all my graphs to a node using the cli just by specifying the hostname. I could have solved this using discovery templates.

2) the important thing I have no answer for
CLUSTER wide aggregates. Yes, I do have the aggregate plugin but I am really limited in that when I add a new host to a cluster I have to hand edit or generate all the aggregates.

I would like to see the total disk usage across a cluster or the average disk usage across a cluster. Any thoughts? This would have to be auto-updating or simple to manage. This is the heart of ganglia start big drill down, cacti works from the bottom up.

Edward

[appodictic]

http://hcc.unl.edu/ganglia/?c=red-worke ... nding&hc=4

[gandalf]

We already discussed such a feature. From my point of view, it's a valid request but simply no time for implementation
R

[TheWitness]

If you get the gmond data into the Cacti database, you can do this quite easily. Also, you could do this using dsstats and a very simple plugin that is aware of your cluster topology (host groups, models, types, etc).

I wrote a nice plugin for LSF for my company. It's pretty cool, and has a pretty good following.

TheWitness

[appodictic]

I do not have a gmond database, just want cacti to collect/display summary data in a similar way.

Here is a simple idea. How easy is it to write an aggregate tool that can work from the CLI? With this i can use the CLI to query for new datasources and then use aggregate_cli to regenerate graphs.