andrea.campi at zephirworks
Mar 26, 2012, 11:40 PM
Post #2 of 3
On Tue, Mar 27, 2012 at 6:13 AM, John Skopis <jspam [at] skopis> wrote:
> I made a daemon to collect stats from varnishstat and calculate 1/5/15
> min averages, and 1/5/15 min averages of the deltas. The stats are
> then exported over HTTP.
> I intend on using the check_varnish.pl script (queries varnishstatd)
> to monitor our varnish instance using nagios.
> The code is here: https://github.com/johnskopis/varnishstatd
> I would love to hear your comments/suggestions.
Reading the name I was hoping you were going in the direction I'm
thinking of going :)
That is, pushing statistics to statsd  and from there to Graphite.
My rationale is that we are going to use statsd for other components
of our architecture; but even looking at just Varnish, it has
potential to be more scalable. In particular, I'm looking at
collecting more in-depth stats via the SHM.
Last year I did a spike on that [2,3] and while my naive approach
broke down under moderate load, it looks like if I keep processing at
a minimum and just fire off UDP I should be able to keep up with a
much higher traffic.
I'm not ready to work on that right now, but I will get there. If this
is something that might interest you, maybe we can combine efforts.
varnish-dev mailing list
varnish-dev [at] varnish-cache