On 03/16/2015 05:40 AM, Eligijus Vitkauskas wrote:
Hello,
I'd like to know where is a proper place to accumulate stats. Imagine
this case:
Server #1:
* Polls memory info every second
* Give this info to StatsAccumulator
* StatsAccumulator outputs every 10s via TCP to Server #2
Server #2 (responsible for displaying monitoring stuff):
* Listens for TCP
* Outputs to Elasticsearch
My question if Stats accumulation should be handled on Server #2 instead
of Server #1. What are the best practices on this?
I'd do the accumulation on Server #1. There's no benefit I can see to
using Server #2, and you're just pushing more data than necessary over
the wire.
Also does it make sense to store memory and cpu usage on elasticsearch
and display graphs with Kibana or should I use tools like Graphite for this?
There's no right answer. Graphite is generally considered to be better
suited for numeric time series data. InfluxDB is a newer popular time
series database. ElasticSearch's sweet spot is more related to text
indexing. People do use ES for time series data, but it's not typically
the tool of choice when the data is purely numeric.
If it turns out that storing mem and cpu information on elasticsearch is
ok than is it ok to store that data on the same index that I store my
other logged information (like http access log)?
I wouldn't.
-r
_______________________________________________
Heka mailing list
[email protected]
https://mail.mozilla.org/listinfo/heka
