You'd better use some already available and mature project like http://www.zabbix.com/, http://www.cacti.net/, http://argus.tcp4me.com/, https://github.com/python-diamond/Diamond or https://github.com/google/grr/ or even ELK (Kibana + Logstash + ElasticSearch).
This project is implemented only as a demo and to test coroutines in Python 3.4
Python version >=3.4.4 is required (for asyncio.ensure_future).
Collector should collect "batches" of metrics like CPU usage, memory usage and disk utilization, and later push them to some external API for actual storage and visualization
Collector uses psutil to get come system stats and saves everything to local database and does not use any external API.
All code is for demo purposes only: storage (tinydb) is slow and unoptimized.
After installing (e.g. with pip install ./async-metrics-collector) you can:
Start collector daemon (using python-daemon that implements PEP 3143 without specifying any configuration options):
mcollector --daemonize --interval=5
Generate plaintext metrics data (usable as input for Graphite):
mcollector --graphite-data
Dump already stored metrics data in Graphite plaintext format:
mcollector --dump