Hi STX team 

This is the follow-up mail to share with the community the test case presented yesterday on the community call meeting. This test case responds to phase 0 on our goal to have a test framework (using existing tools ) to measure performance metrics on StarlingX project. 

This very basic script only has the scope o measure the footprint of STX host system ( users can decide if it is a controller or a node )

This test case was presented yesterday, here are the slides: 

The source code of the test has been updated in tools-contrib repository: 

The list of TODO things for this test case ( and framework ) is here: 

Please feel free to add as many requests as we have as a community.

This test case also add the tools to save the data recorded on an InfluxDB database

The way to run this script as described on the README is: 

python metrics.py

Details of elapsed time and sampling time are described in README as well

By default the script does not send any data, only print on the console the values:

Example Output: https://gist.github.com/VictorRodriguez/b9260ad223b176c323363ac06d0afdde

If we want to send the data to an influxDB: 

python metrics.py --send_data

Will send data to the InlfuxDB server configured in the file: 

$ server.conf

INFLUX_SERVER=<yoruservername or localhost if is the same as the system under test>

INFLUX_PORT=PORT

INFLUX_PASS=PASS

INFLUX_USER=USER

DB_NAME=starlingx

Since yesterday meeting I left the script running on my ubuntu machine and here are the metrics of DB size: 

Size in data points per table:

name: boottime

time count count_1 count_2

---- ----- ------- -------

0 1571

name: cpu_footprint

time count count_1 count_2

---- ----- ------- -------

0 1562

name: hd_footprint

time count count_1 count_2

---- ----- ------- -------

0 3132

name: memory_footprint

time count count_1 count_2

---- ----- ------- -------

0 3132

Size in HD:

du -sh /var/lib/influxdb/data/starlingx/

28K /var/lib/influxdb/data/starlingx/

As we can see in a total of 7827 data points we are only using 28K of the HD ( not much )

Curtis and will be working on the set up of this infrastructure on packet infra, as soon as we have information we will let you know.

This perf framework is designed with the idea that anyone can use it to track their version of STX that you prefer. If you have any question how to set up on your infra please let me know 

Regards

Victor Rodriguez