Skip to content

feat: Add cell-local load balancer health checks - #629

Draft
scotwells wants to merge 3 commits into
codex/nlb-route-reconcilefrom
codex/nlb-health
Draft

scotwells wants to merge 3 commits into
codex/nlb-route-reconcilefrom
codex/nlb-health

Conversation

@scotwells

Copy link
Copy Markdown
Contributor

Summary

Backend health must be verified from a cell network path before routes can send traffic to a service. This adds a cell-local health contract and active TCP, HTTP, and HTTPS checks with bounded work and certificate verification. Consumers receive only fresh results tied to the current route and backend identities, so replacement or stale status cannot keep an endpoint eligible. Checks remain inactive until a VPC-aware socket path is supplied; the contract does not mark traffic ready or announce addresses.

Test plan

  • Verify healthy transitions and immediate removal after a failed probe
  • Verify HTTP responses and HTTPS certificate and IP address validation
  • Verify stale generations, replaced backends, and expired results are excluded
  • Run cell integration and admission checks in CI

Related to #611
Related to #625
Related to #627

Define a health result that a future cell programmer can consume only while
its endpoint identity, route version, and probe evidence remain fresh. Keep
active checks disabled until a VPC-aware socket path is available.

Key changes:
- add a cell-local health API and generated CRD with per-target expiry
- implement bounded TCP, HTTP, and verified HTTPS probes
- reject stale or replaced targets and test probe transitions
Handle probe connection and response cleanup explicitly so CI can validate
the cell-local health contract without lint failures.

Key changes:
- check pipe close errors in focused probe tests
- handle response cleanup and simplify expiry comparison
Bring the health contract onto the updated route foundation while retaining
the published health branch history. Address allocation is now per location.

Key changes:
- merge the route branch's per-location API update
- keep the pull request diff scoped to health work
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant