Optimizing inactive cloud workloads in CrownLabs

This repo is intended to be a quick documentation and report about the work done to implement an inactive resource optimization policy in CrownLabs.

Requirements

• Inactive detection policy (initially based on Tenant last login, could be extended to instance metrics such as workload or ingress/proxy access)
• Tenant notification when an inactive instance is detected (email)
• Instance pause/deletion after max inactive time
• (opt.) Adjust pod resource limits based on the actual consumptions
• (opt.) Live migration for resource consolidation among worker nodes

REPORTS ACHIEVEMENTS

• 31/03 - 06/04: First architecture alternatives
• 07/04 - 13/04: Prometheus-stack installed; Achieved good capability to interact with PromQL, setting custom Alert rules;
• 14/04 - 20/04: Definition and tests of local development environment
• 21/04 - 27/04: Local deployment enviroment testing
• 28/04 - 04/05: Developed a demo implementation of the InstanceInactiveController.
• 05/05 - 11/05: Refined logic for the termination mechanism
• 12/05 - 18/05: Implementation of Prometheus checks for InstanceInactiveController, implementation of mail notification service
• 19/05 - 25/05: Implementation tests
• 26/05 - 01/06: Implementation tests
• 26/05 - 01/06: Implementation of new Grafana Dashboard for Crownlabs resource usage and consumption
• 02/06 - 08/06: Implementation bastion SSH checks
• 09/06 - 15/06: Fixes after Review comments, Refined Grafana Dashboard
• 16/06 - 22/06: Fixes after Review comments, Refined Grafana Dashboard