OSS-Singularity/zcode-dispatch-observatory
OSS-Singularity/zcode-dispatch-observatory
Watch
1
0
Fork
You've already forked zcode-dispatch-observatory
 0
Code Issues Pull requests Projects Releases 137 Packages Wiki Activity Actions

Crash resilience: systemd unit + systemd-aware restart script (v1.25.0) #50

Merged
ClaudiuSchuster merged 1 commit from oss-oo/obs-crisis-hardening-nkyw20260926 into main 2026-09-26 00:17:23 +00:00
Conversation 0 Commits 1 Files changed 6 +199 -3
ClaudiuSchuster commented 2026-09-25 23:51:41 +00:00
Owner

The observatory is a link in the orchestration chain (bells → broker → observatory → owner); its crash during the 2026-09-25/26 triple crisis needed a manual restart. This round adds the safety net without changing the opt-in character:

  • zcode-observatory.service — Restart=on-failure systemd user unit, deliberately NOT auto-enabled (owner preference); install.sh templates the ExecStart path, installs the unit and prints the enable command without running it (--no-unit opts out).
  • observatory-restart.sh — prefers systemctl --user restart when the unit exists; the legacy fallback uses a bracketed pgrep pattern (the ps-grep predecessor killed its own shell on the station) and logs under XDG state (never /tmp, which the station wipes on reboot).
  • managed-parent steer spawns — the send-button and fastlane paths mark their dispatch children so the dispatcher’s v1.0.1 notification scope self-check keeps them in place (the fastlane branches on exit codes, the status view polls liveness — a re-exec would fake deliveries).
  • Tests pin the unit’s restart/opt-in contract, the restart script’s systemd branch / self-kill-safe pattern / log location, and install.sh’s never-enables wiring — 45 unit tests green, CI compile/boot green, Super-Linter green (all six lanes).

Pairs with zcode-dispatch v1.0.1 (PR #31): broker restart policy + watchdog, scoped notification bells, canary infra group whose failures surface in the contract banner.

The observatory is a link in the orchestration chain (bells → broker → observatory → owner); its crash during the 2026-09-25/26 triple crisis needed a manual restart. This round adds the safety net without changing the opt-in character: - **`zcode-observatory.service`** — `Restart=on-failure` systemd user unit, deliberately NOT auto-enabled (owner preference); `install.sh` templates the ExecStart path, installs the unit and prints the enable command without running it (`--no-unit` opts out). - **`observatory-restart.sh`** — prefers `systemctl --user restart` when the unit exists; the legacy fallback uses a bracketed pgrep pattern (the ps-grep predecessor killed its own shell on the station) and logs under XDG state (never `/tmp`, which the station wipes on reboot). - **managed-parent steer spawns** — the send-button and fastlane paths mark their dispatch children so the dispatcher’s v1.0.1 notification scope self-check keeps them in place (the fastlane branches on exit codes, the status view polls liveness — a re-exec would fake deliveries). - Tests pin the unit’s restart/opt-in contract, the restart script’s systemd branch / self-kill-safe pattern / log location, and install.sh’s never-enables wiring — 45 unit tests green, CI compile/boot green, Super-Linter green (all six lanes). Pairs with zcode-dispatch v1.0.1 (PR #31): broker restart policy + watchdog, scoped notification bells, canary infra group whose failures surface in the contract banner.
Crash resilience: systemd unit + systemd-aware restart script (v1.25.0)
All checks were successful
Checks / Functional checks (pull_request) Successful in 16s
Details Explain this check
8c0b5edc61 
The observatory is a link in the orchestration chain (bells → broker →
observatory → owner); its crash during the 2026-09-25/26 triple crisis
needed a manual restart. This round gives it a safety net without
changing its opt-in character:

- zcode-observatory.service: Restart=on-failure unit, deliberately NOT
  auto-enabled (owner preference) — install.sh templats the ExecStart
  path, installs it into ~/.config/systemd/user/ and prints the enable
  command without running it (--no-unit opts out)
- observatory-restart.sh: prefers systemctl --user restart when the
  unit exists; legacy fallback uses a bracketed pgrep pattern (the
  ps-grep predecessor killed its own shell on the station) and logs
  under XDG state (never /tmp, which the station wipes on reboot)
- steer spawns (send button + fastlane) mark their dispatch children
  managed-parent so the dispatcher's v1.0.1 notification scope
  self-check keeps them in place — the fastlane branches on exit codes
  and the status view polls liveness; a re-exec would fake deliveries
- tests: the unit's restart/opt-in contract, the restart script's
  systemd branch, self-kill-safe pattern and log location, install.sh's
  unit wiring (installs, never enables)
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
OSS-Singularity/zcode-dispatch-observatory!50
No description provided.