Apply capacity algorithm changes

* This also adds fields to the instance view for tracking cpu and
  memory usage as well as information on what the capacity ranges are
* Also adds a flag for enabling/disabling instances which removes them
  from all queues and has them stop processing new work
* The capacity is now based almost exclusively on some value relative
  to forks
* capacity_adjustment allows you to commit an instance to a certain
  amount of forks, cpu focused or memory focused
* Each job run adds a single fork overhead (that's the reasoning
  behind the +1)
This commit is contained in:
Matthew Jones
2018-01-11 13:33:35 -05:00
parent 6a85fc38dd
commit 70bf78e29f
17 changed files with 248 additions and 76 deletions

View File

@@ -20,6 +20,8 @@ import six
import psutil
from StringIO import StringIO
from decimal import Decimal
# Decorator
from decorator import decorator
@@ -45,7 +47,7 @@ __all__ = ['get_object_or_400', 'get_object_or_403', 'camelcase_to_underscore',
'ignore_inventory_computed_fields', 'ignore_inventory_group_removal',
'_inventory_updates', 'get_pk_from_dict', 'getattrd', 'NoDefaultProvided',
'get_current_apps', 'set_current_apps', 'OutputEventFilter',
'extract_ansible_vars', 'get_search_fields', 'get_system_task_capacity',
'extract_ansible_vars', 'get_search_fields', 'get_system_task_capacity', 'get_cpu_capacity', 'get_mem_capacity',
'wrap_args_with_proot', 'build_proot_temp_dir', 'check_proot_installed', 'model_to_dict',
'model_instance_diff', 'timestamp_apiformat', 'parse_yaml_or_json', 'RequireDebugTrueOrTest',
'has_model_field_prefetched', 'set_environ', 'IllegalArgumentError', 'get_custom_venv_choices']
@@ -632,19 +634,52 @@ def parse_yaml_or_json(vars_str, silent_failure=True):
return vars_dict
@memoize()
def get_system_task_capacity():
def get_cpu_capacity():
from django.conf import settings
settings_forkcpu = getattr(settings, 'SYSTEM_TASK_FORKS_CPU', None)
env_forkcpu = os.getenv('SYSTEM_TASK_FORKS_CPU', None)
cpu = psutil.cpu_count()
if env_forkcpu:
forkcpu = int(env_forkcpu)
elif settings_forkcpu:
forkcpu = int(settings_forkcpu)
else:
forkcpu = 4
return (cpu, cpu * forkcpu)
def get_mem_capacity():
from django.conf import settings
settings_forkmem = getattr(settings, 'SYSTEM_TASK_FORKS_MEM', None)
env_forkmem = os.getenv('SYSTEM_TASK_FORKS_MEM', None)
if env_forkmem:
forkmem = int(env_forkmem)
elif settings_forkmem:
forkmem = int(settings_forkmem)
else:
forkmem = 100
mem = psutil.virtual_memory().total
return (mem, max(1, ((mem / 1024 / 1024) - 2048) / forkmem))
def get_system_task_capacity(scale=Decimal(1.0)):
'''
Measure system memory and use it as a baseline for determining the system's capacity
'''
from django.conf import settings
if hasattr(settings, 'SYSTEM_TASK_CAPACITY'):
return settings.SYSTEM_TASK_CAPACITY
mem = psutil.virtual_memory()
total_mem_value = mem.total / 1024 / 1024
if total_mem_value <= 2048:
return 50
return 50 + ((total_mem_value / 1024) - 2) * 75
settings_forks = getattr(settings, 'SYSTEM_TASK_FORKS_CAPACITY', None)
env_forks = os.getenv('SYSTEM_TASK_FORKS_CAPACITY', None)
if env_forks:
return int(env_forks)
elif settings_forks:
return int(settings_forks)
_, cpu_cap = get_cpu_capacity()
_, mem_cap = get_mem_capacity()
return min(mem_cap, cpu_cap) + ((max(mem_cap, cpu_cap) - min(mem_cap, cpu_cap)) * scale)
_inventory_updates = threading.local()

View File

@@ -24,7 +24,7 @@ def _add_remove_celery_worker_queues(app, instance, worker_queues, worker_name):
queue['alias'] in settings.AWX_CELERY_QUEUES_STATIC:
continue
if queue['name'] not in ig_names | set([instance.hostname]):
if queue['name'] not in ig_names | set([instance.hostname]) or not instance.enabled:
app.control.cancel_consumer(queue['name'], reply=True, destination=[worker_name])
removed_queues.append(queue['name'])
@@ -43,7 +43,6 @@ def update_celery_worker_routes(instance, conf):
'awx.main.tasks.purge_old_stdout_files',
]
routes_updated = {}
# Instance is, effectively, a controller node
if instance.is_controller():
tasks.append('awx.main.tasks.awx_isolated_heartbeat')