Move scheduler implementation to LAPIS

[eileen-kuehn]

The scheduler implementation is currently contained within this extension although it should be part of the core LAPIS package. However, there are still some hacks that specifically implement caching-related logic to make the scheduler work. Those should be removed before the CondorClassadJobScheduler can actually be moved.

Locations that are caching-specific include:

• details in the WrappedClassAd

lapis.caching/lapis/scheduler.py

Lines 138 to 188 in 8f4f614

	elif "cache_demand" == item:
	caches = self._wrapped.connection.storages.get(
	self._wrapped.sitename, None
	)
	try:
	return mean(
	[1.0 / cache.connection._throughput_scale for cache in caches]
	)
	except TypeError:
	return 0
	elif "cache_scale" == item:
	caches = self._wrapped.connection.storages.get(
	self._wrapped.sitename, None
	)
	try:
	return mean(
	[cache.connection._throughput_scale for cache in caches]
	)
	except TypeError:
	return 0
	elif "cache_throughput_per_core" == item:
	caches = self._wrapped.connection.storages.get(
	self._wrapped.sitename, None
	)
	try:
	return sum(
	[
	cache.connection.throughput / 1000.0 / 1000.0 / 1000.0
	for cache in caches
	]
	) / float(self._wrapped.pool_resources["cores"])
	except TypeError:
	return 0
	elif "cached_data" == item:
	return self._wrapped.cached_data / 1000.0 / 1000.0 / 1000.0
	elif "data_volume" == item:
	return self._wrapped._total_input_data / 1000.0 / 1000.0 / 1000.0
	elif "current_waiting_time" == item:
	return time.now - self._wrapped.queue_date
	elif "failed_matches" == item:
	return self._wrapped.failed_matches
	elif "jobs_with_cached_data" == item:
	return self._wrapped.jobs_with_cached_data

• specifics in the matching itself

lapis.caching/lapis/scheduler.py

Lines 915 to 922 in 8f4f614

	# monitoring/coordination stuff
	if (
	candidate_job._wrapped._total_input_data
	and matched_drone._wrapped.cached_data
	):
	candidate_job._wrapped._cached_data = (
	matched_drone._wrapped.cached_data
	)