167 - Wait for aborted process exit before finalising schedule

[louis-pie]

Related to AP-2830

Context

Cicada currently clears is_running and finalizes the schedule log immediately after abort_running sends SIGTERM, even though the launched process may still be running.

Refer to 0.10.4 CHANGELOG for full details.

Changes

• Keep a schedule marked as running until its launched process has stopped.
• Clear repeated abort requests while a process is stopping so they cannot affect the next run.
• Continue supervising the process through temporary process-check and database failures.
• Preserve the confirmed process return code and safely record diagnostic details.
• Clear execution state and finalize the schedule log atomically after process exit.

Test coverage

• Existing: 2 cases related to abort_running, 5 covering broader exec_schedule behaviour.
• New: 6 cases directly covering abort_running, plus 15 supporting shutdown, error-handling, and database cases.
• Updated: 1 existing abort_running case. It previously checked only the final log result; it now also confirms is_running and abort_running are cleared after the process exits.
• Deleted: None.

Validation

• make pytest: 162 passed, 79.72% coverage
• make flake8: passed with 0 errors
• make black: passed

Checklist

• Change meets or does not compromise the Baseline Security Requirements

[platon-github-app-production]

Comment /request-review to automatically request reviews from the following teams:

• analytics-platform: notifications will be sent to #analytics-platform-public

You can also request review from a specific team by commenting /request-review team-name, or you can add a description with --notes "<message>"

💡 If you see something that doesn't look right, check the configuration guide.

[wise-github-bot-app]

This is currently a non-blocking check, and it will not prevent your Pull Request from being merged.

The CODEOWNER of this repository has some untagged assets.

As this is a public repository, asset details will not be shown.

[Copilot]

Pull request overview

Changes: Bugfix (1), Test improvement (1), Documentation update (1), Meta (1)

This PR updates exec_schedule to keep schedules marked as running until an aborted child process is confirmed exited, and to consume repeated abort_running requests during shutdown waits.

Changes:

• Add consume_abort_running() and update the abort flow to wait for child exit before clearing is_running.
• Add focused unit tests covering abort supervision and state transitions.
• Update docs/config examples (YAML quoting for "*", abort semantics, ERD source link) and refresh repo guidance/changelog entries.

Reviewed changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
cicada/commands/exec_schedule.py	Implements the “assume running until polled closed” behavior and consumes repeated abort requests while waiting.
tests/test_exec_schedule.py	New unit tests validating abort wait semantics, repeated abort consumption, and state-clearing SQL.
tests/test_functional_main.py	Fixes YAML for returncodes_alert and resets the Slack mock between loop iterations.
config/example.yml	Updates example config to quote "*" so it parses as a string instead of YAML alias syntax.
README.md	Documents updated abort behavior and links the editable ERD source.
docs/erd.excalidraw	Adds the editable Excalidraw ERD source file.
CLAUDE.md	Replaces prior content with authoritative repository guidance and workflow expectations.
CHANGELOG.md	Adds a new 0.10.3 section describing the abort-running behavior and ERD source addition.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 23 out of 25 changed files in this pull request and generated no new comments.

Suppressed comments (2)

.github/workflows/linter.yml:19

• This job uses a GitHub-hosted runner, which violates the organization requirement that all workflow jobs run on self-hosted gha-production-* runners. The lint workload should use the standard production runner.

    runs-on: ubuntu-22.04

.github/workflows/pytest.yml:20

• This job uses a GitHub-hosted runner, which violates the organization requirement that all workflow jobs run on self-hosted gha-production-* runners. Use the standard medium runner (or a larger approved tier if the Docker test workload requires it).

    runs-on: ubuntu-22.04

[platon-github-app-production]

Comment /request-review to automatically request reviews from the following teams:

• analytics-platform: notifications will be sent to #analytics-platform-public

You can also request review from a specific team by commenting /request-review team-name, or you can add a description with --notes "<message>"

💡 If you see something that doesn't look right, check the configuration guide.

[Copilot]

Pull request overview

Copilot reviewed 10 out of 11 changed files in this pull request and generated no new comments.

Suppressed comments (1)

cicada/commands/exec_schedule.py:283

• Deliver the shutdown signal before checking abort_running. As written, a received SIGTERM/SIGQUIT is recorded here, but the next operation opens a database connection; because db_cicada() has no connect_timeout, a database outage can delay child termination for the connection timeout plus the retry sleep. This leaves the launched process running during supervisor shutdown. Terminate immediately once the shutdown request is observed, and retry a failed delivery before performing the database check.

            signal_number = shutdown_request["signal"]
            if signal_number is not None and requested_stop_result is None:
                requested_stop_result = ExecutionResult(-15, f"{signal.Signals(signal_number).name} received")