Devilator - performance monitoring for FreeBSD

[Borja Marcos]

> This is very cool. :-)  How are you currently extracting the 
> information?
> One of the things I've wanted to do for a while is make sure all this 
> sort
> of thing is exposed via snmpd so that the information can be gathered
> easily across a large number of hosts (say, 10,000).

	Right now I'm using a combination of sysctl() and kvm accesses. The 
system is simple: a small agent reads those values and places them in a 
text file. That text file is then processed by the Orca program.

	You can set it up so that each machine only runs the small agent, and 
you fetch the text files with (for example) rsync from another machine 
dedicated to generating the graphs and serving the web pages for all 
your hosts. I'm right now using a Sun Netra T1 running FreeBSD 5.3 and 
it's generating pages for all of our Solaris hosts.

	The advantage of this approach is security. It's much easier to secure 
an rsync access than snmp, and the data collection is much more 
precise.

> systat contains a large number of useful statistics and monitoring 
> pieces,
> including some really useful stuff in systat -vmstat.  There's also a 
> lot
> of interesting information to be had from vmstat -z and vmstat -m in 
> terms
> of tracking system resource use.

	Yes, I've been having a look at the systat and vmstat source code.

> It would be neat to have interrupt rates classified by source, so that 
> the
> regularity of clock interrupts was visible against the variation in 
> things
> like disk and network interrupts.

	That's in my list in one way or another.




	Borja.