[collectd] customizing metrics paths for various plugins

Florin Andrei florin at andrei.myip.org
Sun Mar 8 22:05:47 CET 2015

For a previous project, I've used Sensu to collect metrics - more 
specifically, sensu-client as an agent. That had the side effect of also 
providing fault-detection and alerting via Sensu. The whole thing worked 
great. The paths for the collected metrics were very flexible. Any Sensu 
plugin could use arbitrarily specified metric paths when returning data.

That freedom is fantastic. It allows you to structure the data any way 
you want in the datastore. Queries become simple and logical. Setting up 
a Grafana dashboard is easy. I think being able to specify the metric 
path for each metric, or at least for each plugin, is becoming a crucial 
feature these days.

OTOH I'd like to use collectd for this project, because of the 
simplicity of deploying agents, and the promise of high performance. I 
might be willing to hack the source code and rebuild the packages in 
order to change the data paths, if there is no other way to do it.

On 2015-03-06 20:42, Stuart Cracraft wrote:
> I need this also.
> We use pingdom and have a large number of externally driven
> metrics which we API out to and incoming from our network.
> collectd is gathered and the charts are all flat and take up a lot
> of space in graphite.
> Anyone have any thoughts on InfluxDB and Grafana?
>> On Mar 6, 2015, at 6:28 PM, Florin Andrei <florin at andrei.myip.org> 
>> wrote:
>> New to this project, I am testing collectd as a metrics collector to 
>> be used with Graphite. Everything's fine so far, and I am very 
>> impressed with how easy it is to collect high-frequency samples with 
>> very low CPU usage.
>> I've enabled the CPU plugin and the metrics are showing up in 
>> Graphite. But each individual metric is prefixed with "cpu-":
>> collectdlocalhostcollectd/
>> ├── cpu-0
>> │   ├── cpu-idle.wsp
>> │   ├── cpu-interrupt.wsp
>> │   ├── cpu-nice.wsp
>> │   ├── cpu-softirq.wsp
>> │   ├── cpu-steal.wsp
>> │   ├── cpu-system.wsp
>> │   ├── cpu-user.wsp
>> │   └── cpu-wait.wsp
>> This seems redundant, and when visualized with Grafana, it takes too 
>> much space in the legend. Is there a way to get rid of the "cpu-" 
>> prefix in the metrics names? Let's say for the user metric, I want the 
>> path to be...
>> host.cpu-0.user
>> ...instead of...
>> host.cpu-0.cpu-user
>> Same question for the memory plugin.
>> For the load plugin, the path is load.load.longterm - here the load 
>> label appears on two different levels. Can I get rid of one level?
>> Also, is there a way to group some plugins under the same branch? I'd 
>> like CPU, memory, load, etc to fall under the category "os". Other 
>> plugins, such as Apache, etc. I'd like to group under the "app" 
>> category.
>> host
>>  os
>>    cpu
>>    memory
>>    load
>>  app
>>    apache
>> It's not clear to me how to manipulate these paths with collectd.
>> --
>> Florin Andrei
>> http://florin.myip.org/
>> _______________________________________________
>> collectd mailing list
>> collectd at verplant.org
>> http://mailman.verplant.org/listinfo/collectd

Florin Andrei

More information about the collectd mailing list