README.md

SLO

You'll find some proof of concepts, research documents, es queries...

Generate latency logs

Generates some fake latency logs that can be used for creating SLOs on. The logs contains latency field.

SCENARIO can be either:

• 90percent_good: 90% of the time, the latency is below 100ms ([0, 100]), or above 300ms ([300, 600])
• 95percent_good: 95% of the time, the latency is below 100ms ([0, 100]), or above 300ms ([300, 600])
• 99percent_good: 99% of the time, the latency is below 100ms ([0, 100]), or above 300ms ([300, 600])
• full_outage_every_day: Follow a sinusoidale function to generate the good (f(x) >= 0) and bad (f(x) < 0) logs. Function is meant to be negative once per day-ish
• healthy_then_failing: Previous data is 100% good, then starts failing

	SCENARIO="95percent_good" node lib/latency_logs_generator/index

Generate SLO historical data and continous data

Note: this script is deprecated. It simulates the output of an SLO's running transform. Use the latency logs generator instead.

Handy if you want to generate some data for an existing SLO, and keep generating the data as time passes. The script will generate 3 month of historical data, and then every minute, will insert a new document.

1. Create an SLO through the API. Copy the slo id
2. Stop the transform manually associated to this slo
3. Remove the data manually (delete by query or delete all slo indices)
4. Run node lib/continuous_rollup_data_generator/index.js slo_id

The data is generated with the following noise function. When the function returns a negative value, we use it to compute the good events as a ratio of the total events.

Generate historical summary data

Generates some historical summary data to be use in tests/storybook. Tweak the scripts under lib/historical_summary_generator to match the expected use case then run:

  node lib/historical_summary_generator/index