[Epic] Testing Overhaul 2025

[dereknola]

This is an epic to track all the changes for testing in 2025. Since most testing is self tested in the CI, QA for most of this is not necessary. This issue largely exists as an organizational and historical record.

Docker Tests

• Convert legacy docker tests from bash to Golang #11419
• Convert all Drone pipelines to use new Golang Docker Tests #11420
• Investigate replacement of Sonobuoy conformance with Hydrophone
• Investigate use of rancher/systemd-node as option to allow restarts of k3s

E2E Tests

• Investigate use GHA self-hosted large runners to improve performance of k3s build job.

  • See Try running e2e on big runners #11355 for full notes, but ultimately, the lack of availability on the equinix runners (only 1 instance at a time) destroyed the benefit of faster builds (7m vs 5m)

• Investigate use of compiling go tests for increase performance

  • Performance gains were nonexistent, either margin-of-error, or worse performance. See Compiled vs Regular.

• After docker migrations, determine which e2e tests should be "downgraded" to docker tests

• Upload systemctl node logs as artifacts in GHA on failure

• Increase Utilization of amd64 Drone Runner #11449

Integration Tests

• Investigate use of compiling go tests for increase performance
  • Performance gains were slim, usually under ~20s difference. See Compiled vs Regular.

Unit Tests

• Overhaul old loadbalancer unit test (Completed with Rework loadbalancer server selection logic #11329)

General Testing

• Increase code coverage across K3s #11484