[RUN-3009] implements kill on abort

[ronaveva]

This pull request introduces robust process tree termination for WinRM-executed jobs on Windows nodes, addressing the long-standing issue where aborting a job left remote processes running. The main feature is an optional "Terminate On Abort" mode, which—when enabled—ensures that aborting a job kills both the remote command and all its child processes. This is achieved by capturing the remote process PID and force-killing the process tree on abort. The implementation includes new code for PID capture, output filtering, and remote termination, as well as user-facing documentation and configuration options.

The most important changes are:

Process Tree Termination Feature:

• Added a "Terminate On Abort" option to the plugin configuration (plugin.yaml, README.md). When enabled, aborting a job will terminate the remote command and its entire process tree on the Windows node. This is disabled by default for backward compatibility. [1] [2] [3]

Implementation of Remote PID Capture and Termination:

• Introduced a new module winrm_kill.py that provides:
  • A preamble to print the remote shell's PID,
  • Functions to parse and strip PID marker lines from output,
  • A MarkerFilter class for streaming output filtering,
  • Logic to send both the WS-Man terminate signal and a taskkill command to kill the entire process tree on abort.

Integration with Job Execution Flow:

• Modified winrm-exec.py to:
  • Detect the "Terminate On Abort" setting and wrap commands with the PID-capturing preamble,
  • Install signal handlers to catch aborts and trigger remote termination,
  • Filter output to hide PID marker lines and capture the PID,
  • Ensure clean-up and forced exit on abort. [1] [2] [3] [4] [5]

WinRM Session Enhancements:

• Updated winrm_session.py to:
  • Track live WinRM handles and the remote PID for each running command,
  • Allow abort handlers to access these handles and perform remote termination,
  • Pass a tracker object through command execution to enable this coordination. [1] [2] [3] [4]

Documentation:

• Expanded the README.md with a new "Aborting jobs" section, explaining the new behavior, configuration, and technical details for users and operators.

These changes collectively ensure that aborting a WinRM-executed job on Windows nodes can now reliably clean up all spawned processes, improving reliability and resource management.

[Copilot]

Pull request overview

This PR adds an optional “Terminate On Abort” mode to the WinRM node executor so that aborting a Rundeck job can terminate the remote WinRM command and its entire child process tree (via PID capture + taskkill /F /T), addressing orphaned processes on Windows nodes.

Changes:

• Added remote PID capture + streaming output marker filtering and an abort-time termination routine (terminate_remote) that combines WS-Man terminate + taskkill /T.
• Wired abort handling into the WinRM execution flow (signal handling, PID propagation via a tracker object).
• Added user-facing configuration and documentation plus unit tests for the kill helpers.

Reviewed changes

Copilot reviewed 7 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
contents/winrm_kill.py	New PID marker/preamble, output filtering, and remote termination orchestration.
contents/winrm-exec.py	Integrates terminate-on-abort behavior (signal handling, wrapping commands, filtering marker output).
contents/winrm_session.py	Exposes live WinRM handles (shell/command IDs) to support abort-time termination.
plugin.yaml	Adds the new terminateonabort configuration option.
README.md	Documents “Terminate On Abort” behavior and operational notes.
tests/test_winrm_kill.py	Unit tests for PID parsing/stripping, streaming filter, and terminate orchestration.
contents/common.py	Minor regex literal cleanup (raw strings).
.gitignore	Ignores common Python build/test artifacts.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.