fix!: concurrent worker agent processes

[jusiskin]

What was the problem/requirement? (What/Why)

The PosixInstanceWorkerBase class, which forms the base class of PosixInstanceBuildWorker had a bug causing multiple worker agent processes to be running simultaneously.

There was code in one location that was starting the systemd unit for the worker agent:

deadline-cloud-test-fixtures/src/deadline_test_fixtures/deadline/worker.py

Lines 784 to 785 in 84592ad

	if config.start_service:
	cmds.append("systemctl start deadline-worker")

and code in another place launching the worker agent in an SSM command:

deadline-cloud-test-fixtures/src/deadline_test_fixtures/deadline/worker.py

Line 644 in 84592ad

f"nohup runuser --login {self.configuration.agent_user} -c 'AWS_DEFAULT_REGION={self.configuration.region} deadline-worker-agent > /tmp/worker-agent-stdout.txt 2>&1 &'",

What was the solution? (How)

Removed the code that launches the worker agent through an SSM command, since the officially supported and recommended way to run the worker agent is using systemd. We want tests to reflect this setup.

The way configuration is applied to the worker agent had to be changed. We were previously setting environment variables in the worker agent user's bash startup script, but this script is not loaded by systemd services since it does not use a shell. Instead we use a supplementary systemd drop-in config file (see https://wiki.archlinux.org/title/Systemd#Drop-in_files)

I've also modified the default for DeadlineWorkerConfiguration.start_service from false → true since most tests relied on the worker being started. Downstream consumers that relied on the worker agent being started by the SSM command will not need to modify their test code and this is the sensible default.

What is the impact of this change?

There will not be issues encountered in tests using this library where there are multiple worker agents running for the same worker ID, causing conflicts.

How was this change tested?

1. Checked out the mainline branch of https://github.com/aws-deadline/deadline-cloud-worker-agent

2. Created the hatch environments (e.g. ran hatch run fmt)

3. Installed the modified version of dealdine-cloud-test-fixtures from this branch using:

   hatch shell
   pip install -e ../deadline-cloud-test-fixtures
   

4. Ran the E2E tests using the instructions from DEVELOPMENT.md

Was this change documented?

No

Is this a breaking change?

Yes:

1. The default of DeadlineWorkerConfiguration.start_service was changed from False → True
2. Downstream consumers of this package may have relied on the worker agent being started through the SSM command.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[sonarcloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarCloud