Archived community.zenoss.org | full text search
Skip navigation
7311 Views 3 Replies Latest reply: Jun 15, 2010 11:47 AM by Matt Ray RSS
jamesmc Newbie 2 posts since
Jun 15, 2010
Currently Being Moderated

Jun 15, 2010 6:41 AM

Performance metrics: Multiple values for the same data point?

Hiya,

 

I'm new to Zenoss, and evaluating the core edition currently for the purpose of gathering performance data.

(I've gone through a fair few threads/articles about adjusting rrds, cycle times, graph layouts, command data sources etc and it's progressing nicely)

 

However, I'm having trouble gathering performance data where I'd like to return multiple data points _for the same metric_ at a time.

(The documentation mentions in a few areas about this typically handled in a 'one data point per metric' fashion - which I have working)

The data I'm after will be down to second granularity, I think too much for a standard snmp data source to cope with.

 

The Nagios docs seem to say that _different_ data points can all be returned, using a 'space separated list of label/value pairs'.

I have a check_ping type example in place already that works with ssh and graphs a data point for 'rta'.

I did check http://nagiosplug.sourceforge.net/developer-guidelines.html but may have missed something .. ?

 

Just to take this one as an example, if I want to run 'check_ping' via ssh every minute. (This is just for the sake of discussion!)

Can I have it return (the last?) 60 data points, (one for each second)? - How is something like this handled?

(I can make adjustments to what is being fed to stdout on the remote end, return codes etc no issue there)

Can RRD handle graphing these data points in 'batches' like this over time?

How much data can be fairly reliably passed back each time in this way ?

 

At the moment I have something like this returned:

OK ... |rta=<value>

Maybe can the same label just be repeated? (Link above seems to say the labels must be unique)

OK ... |rta=<value>;rta=<value>;rta=<value>;rta=<value> ... (60 times) ?

 

I'm thinking the only way to handle this is to create a custom performance daemon/collector.

Does anyone have a similar situation or suggestions/recommendations about how to handle a situation like this ?

Something that ties in with the above example would be great

 

Any and all feedback appeciated !

More Like This

  • Retrieving data ...

Legend

  • Correct Answers - 4 points
  • Helpful Answers - 2 points