Skip to content

Fix station bootstrap and defaults; harden application monitoring readiness - #1

Merged
leveldorado merged 7 commits into
mainfrom
codex/doers-monitoring-hardening
Sep 19, 2026
Merged

leveldorado merged 7 commits into
mainfrom
codex/doers-monitoring-hardening

Conversation

@leveldorado

@leveldorado leveldorado commented Sep 19, 2026 •

Copy link
Copy Markdown
Member

A fresh station could fail before CA publication because Zig 0.16 opened the operation-lock directory with Linux O_PATH, which flock rejects even when the lock is free. Open a readable directory descriptor, preserve exclusive nonblocking locking, and distinguish OperationBusy from fixed filesystem/PKI diagnostics. Converge absent or empty ingress directories, recover proven unpublished candidates, reuse published CA/server identity, and restrict registry migration to the known 0700-to-0750 generation. Invalid active state still fails without rotation or private output.

Station install, verify, status and notify-test now load optional CWD ./station.toml. Explicit --config replaces the default, CLI values take precedence, and canonical [station].hostname supplies the independent mTLS identity. The v1 [ingress].hostname alias remains compatible unless contradictory. Plans show connection/hostname/ports without resolving secrets; application commands retain ./monitoring.toml.

The earlier application-hardening changes accept fresh failed-scrape telemetry when the app is down, preserve real-payload requirements for successful targets, improve safe application checks, and read TLS client-auth alerts before sending HTTP. The pinned Doers process fixture covers zero-client station readiness, scoped app publication/removal, trusted metrics/logs, quiet streams, alert hold/recovery, no-op publication and bounded outage buffering. Station/application ownership, service units, rule definitions, pins and public networking scope stay unchanged.

Validation at d8393b7:

  • Zig 0.16.0: zig fmt --check build.zig src, zig build, zig build -Doptimize=ReleaseSafe, and zig build test --summary all passed locally; 398/398 Zig tests (including 45 native lifecycle/PKI tests).
  • python3 tests/cli_smoke.py: 293 checks passed; unavailable Fish checks skipped. Vendor integrity, version/diagnostic and release packaging checks passed.
  • Cross-compiled native lifecycle/PKI binary: 45/45 passed on isolated Ubuntu 24.04 arm64.
  • Real native helper chroot fixture: 10/10 passed on both Ubuntu 24.04 and 26.04, including exact root/dt-ingest empty production state, hostname validation, no-op, lock contention, invalid CA preservation and metadata refusal. No external crypto executable is present in the fixture root.
  • All six GitHub Actions jobs passed at d8393b7: https://github.com/BaptizedDragon/dragontools/actions/runs/35428036279. macOS: 398/398 Zig tests and 293 CLI checks. Linux: 397 passed with one optional shell test skipped, plus 276 CLI checks; absent Zsh/Fish checks were skipped. Both platforms passed pinned Caddy mTLS and release checks; all four ReleaseSafe cross-build/package jobs passed.
  • Linux CI also passed the entire Doers signal/alert/outage fixture, native helper install lifecycle, and all 10 exact-ownership bootstrap process tests. It observed real alert firing/resolution, scoped no-op publication, bounded Vector backpressure, and resumed delivery without an agent restart.

Disposable-host integration not run. No production SSH, service stop, DNS/firewall change or notification was performed. Filesystem/chroot, fake-remote and isolated process tests do not prove real systemd/journald, public listener reachability, two-host deployment or Telegram delivery. Deployment validation remains with the operator.

@leveldorado leveldorado changed the title Harden first application monitoring readiness and Doers reference integration Fix station bootstrap and defaults; harden application monitoring readiness Sep 19, 2026
@leveldorado
leveldorado marked this pull request as ready for review September 19, 2026 07:21
@leveldorado
leveldorado merged commit de54c30 into main Sep 19, 2026
12 checks passed
@leveldorado
leveldorado deleted the codex/doers-monitoring-hardening branch September 21, 2026 07:45
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