Try to avoid workers with the same name.

[vdbergh]

During request_task, we check if there are active tasks for workers with the same name, which have recently been updated (i.e. they are not dead). If so then we return an error.

Should fix #1360 (comment) .

[vdbergh]

Hmm CI fails with

======================================================================
ERROR: test_auto_purge_runs (test_api.TestRunFinished.test_auto_purge_runs)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/runner/work/fishtest/fishtest/server/tests/test_api.py", line 527, in test_auto_purge_runs
    self.assertEqual(response["run"]["_id"], str(run["_id"]))
                     ~~~~~~~~^^^^^^^
KeyError: 'run'

I did not touch autopurge...

[ppigazzini]

I restarted a couple of times the CI, same error, the raw log doesn't show further information.

[ppigazzini]

From the error (missing key), it seems that the API request failed.
The PR changes the code of def sync_request_task(self, worker_info): and that function is called over and over in test_api.py.

fishtest/server/tests/test_api.py

Lines 508 to 518 in 7363126

	def test_auto_purge_runs(self):
	stop_all_runs(self)
	run_id = new_run(self)
	run = self.rundb.get_run(run_id)
	num_games = 1200
	run["args"]["num_games"] = num_games
	self.rundb.buffer(run, True)
	# Request task 1 of 2
	request = self.correct_password_request()
	response = ApiView(request).request_task()

fishtest/server/fishtest/api.py

Lines 426 to 430 in 940e460

	@view_config(route_name="api_request_task")
	def request_task(self):
	self.validate_request("/api/request_task")
	worker_info = self.worker_info()
	result = self.request.rundb.request_task(worker_info)

fishtest/server/fishtest/rundb.py

Lines 689 to 693 in 940e460

	def request_task(self, worker_info):
	if self.task_semaphore.acquire(False):
	try:
	with self.task_lock:
	return self.sync_request_task(worker_info)

[vdbergh]

I see. With this PR a worker can no longer have two active tasks at the same time (roughly). I guess this condition will not be satisfied during some tests. I will look at it tonight.

[vdbergh]

Fixed!

[ppigazzini]

I will look at it tonight.

I wonder what is your timezone right now..

[vdbergh]

I couldn't resist to have a quick look and it turned out to be easy to fix (just changing the order of some statements).

[ppigazzini]

Right now DEV is running with the time aware datetime PR, I will test this PR this evening after loading a clean DB backup.

[ppigazzini]

Looks good on DEV, tested with a couple of workers started with python3 worker.py --uuid_prefix 123456

[ppigazzini]

PROD updated, thank you @vdbergh :)