graphite-dev team mailing list archive

[Florin Andrei <florin@xxxxxxxxxxxxxxx>]

I am thinking to build upon a classic, full blown, vanilla Sensu mesh 
(Sensu client as data collector, RabbitMQ as message broker, Sensu 
server for decision making, Redis for storage), and simply plug into it 
a Graphite instance, via something like WizardVan maybe.

The alternative would be a parallel Graphite deployment straight out of 
the documentation, using collectd as an agent. But I don't want to 
collect the same data twice: one for metrics (Graphite), another for 
alerts (Sensu).

What are the potential issues with plugging Graphite into a complete 
Sensu mesh? Do I lose time granularity? Do I lose anything else? 
Compatibility issues?

How about performance and scalability? I'm sure for a small setup any 
solution would be fine, but if I need to scale up to many instances then 
which architecture would scale better?

(All of the above in the context of probably using something like 
Logstash in the future, so interoperability of all components is important.)

I guess I'm talking about a pretty open-ended scenario here, so any 
observations are welcome.

--
Florin Andrei
http://florin.myip.org/