graphite-dev team mailing list archive
-
graphite-dev team
-
Mailing list archive
-
Message #04215
[Question #231736]: System statistics with min/max aggregates
New question #231736 on Graphite:
https://answers.launchpad.net/graphite/+question/231736
I'm currently deploying Graphite for the first time, moving from RRDTool.
First, let me ask if my understanding is correct:
1) In RRDTool I can feed a single value per RRD and RRDTool would update one or more RRAs, taking care of any aggregation functions configured on its own.
2) In Graphite, even if I configure e.g. last/min/max metrics with proper retentions and aggregation methods, the metrics are pretty much independent from each other, and I have to feed graphite with input data for each of them separately, i.e. I need to pre-aggregate the data (over update interval) and feed Carbon-cache with the pre-aggregated values for avg/min/max metrics.
Do I properly understand the mechanics of things?
The root cause of my question is the difficulty encountered while trying to configure usual system statistics - RAM, load average, CPU, etc. I decided to try collectd, as it seems lightweight, robust and actively developed. However collectd only outputs a single metric per monitored value, e.g. 'collectd.myhost.df-root.df_complex-used' for root fs usage.
If I want to aggregate min/max value over longer periods for historic data, and my understanding is correct, I should supply 3 values to Carbon on each update - for average/min/max metrics. Actually I'd be happy with the same value 3 times, as "long-term min/max aggregate of average" is all I need.
Is it possible with collectd? Is there any collectd replacement that would help? Obviously it is no rocket science and I could easily develop a script/program feeding Carbon with all those simple statistics I want, this however sounds like reinventing the wheel.
--
You received this question notification because you are a member of
graphite-dev, which is an answer contact for Graphite.