Combination graph question - I apologize in advance!

[sg]

While I'm sure this is a much-covered question, I can't find an answer in any of the forums -- so I'm going ahead and asking anyway...

I've got some hosts I'm monitoring, and some nice graphs going. CPU utilization, memory utilization, etc. These hosts are part of a small "cluster".

What I'd really like to do is a summary-type graph, with all the nodes' CPU utilization showing on one graph. Preferably as a cumulative number, but perhaps individually would be okay.

As I'm alreay collecting this, I don't want to create a new datasource thing. There are eight nodes locally, and another eight across the WAN, and I don't want a nasty phonecall from the network administrators.

Any suggestions?

sg

[Lux]

Try this thread. http://www.raxnet.net/board/viewtopic.p ... highlight=

[sg]

This thread was helpful.

At first I had a lot of trouble getting the data from different hosts -- every time I picked a data source from a new source, all the data sources normalized to that new host.

I solved the problem by not using a graph template.

I don't know, however, if I'm using the STACK type properly.

I want to add all the datasources together to produce a total utilization graph. I picked AREA for the first item, then STACK for the other seven. Is this the correct approach?

sg

[Lux]

That should be correct.

Templates should work just fine. I'm doing something similar with some Nortel router/switches and I'm happily using a template for them. Are you trying to associate your graph template with a host when you create the new graph?

Mike

[sg]

I was leaving the host entry as "none" (blank did not seem to be an option).

Other than that, it went well. Not really difficult, once I understood what I was supposed to do.

sg

[Guest]

looking for something like this?
http://monitor.idleserv.net/traffic.html

[Guest]

This is what I was trying to do, yes, but with CPU util vs. traffic.

Interestingly, you have a similar issue to what I'm currently battling.

Your hosts have a bizarre value of over 400 m (400 million?) in your CPU Nice values, but nowhere else.

I tried using the killspike2 script, but I either get the spike down to 14 m, or eliminate all my good values. I can't fiigure it out...

[melchandra]

I think 400 m referes to micro, not million. million would be a upper case M.

[Kybber]

m=milli, u=micro

[sg]

Well, that makes sense. Duh.

[melchandra]

m=milli, u=micro :)

*smacks self*

I knew that, just spazed out. thanks Kybber