fix(retrain): clear recent alerts data and set appropriate last_anomaly_timestamp - Githubissues

chaos-genius / chaos_genius

ML powered analytics engine for outlier detection and root cause analysis.

https://www.chaosgenius.io

MIT License

737 stars 83 forks source link

fix(retrain): clear recent alerts data and set appropriate last_anomaly_timestamp #1128

Closed Samyak2 closed 2 years ago

Samyak2 commented 2 years ago

Changes

Before re-training, the triggered alerts data from T-offset is deleted and the last_anomaly_timestamp is set to one unit lesser than T-offset (unit depends on the anomaly frequency).

Notes for reviewers

See new functions: clear_triggered_alerts_from_offset, set_last_anomaly_timestamp_to_offset
See new code in kpi_anomaly_retraining in anomaly_data_view

Testing

[x] Daily KPI - lyft - individual alert. Alert gets triggered everytime re-training is done. The alert has data for T-offset only.
[x] Hourly KPI - cloud cost - individual alert. Alert gets triggered everytime re-training is done. The alert has data for all hours of today.

gitpod-io[bot] commented 2 years ago

netlify[bot] commented 2 years ago

Deploy Preview for frontend-sb canceled.

Name	Link
Latest commit	2263deed89620bc513b29929dc5bbb1385bdd405
Latest deploy log	https://app.netlify.com/sites/frontend-sb/deploys/632af943722d0700092e4694