[Zenoss-dev] Collecting Application level performance data

[dsudell]

I'm looking at using Zenoss to do performance monitoring of an application stack.  All the OS, Network, etc stuff is pretty well covered.  Instrumenting the application is, of course, my problem.  (Its my application).  But I'm fishing for a bit of advice on how best to do this and pitfalls to avoid.

I could try to use SNMP to monitor the application, but doing a Sub-Agent is less than inviting. 

The application in question is an n-tier application done in python with the tiers communicating via xml-rpc.  So that seemed like a really good angle.

The how to http://www.zenoss.com/community/docs/howtos/data-collector-plugins/
was a little thin, of course its a work in progress.  The developer guide is a bit better http://www.zenoss.com/community/docs/zenoss-dev-guide/2.1/ch07.html, but still makes it feel like uncharted territory.  The one Zen Pack that was clearly a collector (ZenJMXCollector) is about the sort of thing I was thinking, but as mostly Java adds a bit of complexity as an example.

I could fall back to rolling a command line app that polls my apps and sit that under the Command collector.  That clearly works.  Though its a bit less integrated and having to fork all the time scales less well as the application grows.

Am I making this too hard by thinking of building my own collector and avoiding the Command collector?  Is there a good example to look at?  Am I correct in thinking that it requires a daemon? (as contrast with the HttpCollector that simply leverages the CommandCollector by execing check_http as opposed to having used something like urllib directly in python to do the same thing)




-------------------- m2f --------------------

Read this topic online here:
http://community.zenoss.com/forums/viewtopic.php?p=18388#18388

-------------------- m2f --------------------