External-Mirrors/awx
2
0
Fork 0
mirror of https://github.com/ansible/awx.git synced 2026-02-26 15:36:04 -03:30
Code Issues Packages Projects Releases Wiki Activity

Remove committed_capacity field, delete supporting code (#12086)

Browse Source 
* Remove committed_capacity field, delete supporting code

* Track consumed capacity to solve the negatives problem

* Use more verbose name for IG queryset
This commit is contained in:
Alan Rominger
2022-04-22 13:41:32 -04:00
committed by GitHub
parent c43424ed09
commit cb63d92bbf
8 changed files with 100 additions and 191 deletions
Download Patch File Download Diff File Expand all files Collapse all files

119
awx/main/managers.py
View File

@@ -7,10 +7,9 @@ from django.conf import settings
from django.db.models.functions import Lower
from awx.main.utils.filters import SmartFilter
from awx.main.utils.pglock import advisory_lock
from awx.main.utils.common import get_capacity_type
from awx.main.constants import RECEPTOR_PENDING
___all__ = ['HostManager', 'InstanceManager', 'InstanceGroupManager', 'DeferJobCreatedManager', 'UUID_DEFAULT']
___all__ = ['HostManager', 'InstanceManager', 'DeferJobCreatedManager', 'UUID_DEFAULT']
logger = logging.getLogger('awx.main.managers')
UUID_DEFAULT = '00000000-0000-0000-0000-000000000000'
@@ -162,119 +161,3 @@ class InstanceManager(models.Manager):
create_defaults['version'] = RECEPTOR_PENDING
instance = self.create(hostname=hostname, ip_address=ip_address, node_type=node_type, **create_defaults, **uuid_option)
return (True, instance)
class InstanceGroupManager(models.Manager):
"""A custom manager class for the Instance model.
Used for global capacity calculations
"""
def capacity_mapping(self, qs=None):
"""
Another entry-point to Instance manager method by same name
"""
if qs is None:
qs = self.all().prefetch_related('instances')
instance_ig_mapping = {}
ig_instance_mapping = {}
# Create dictionaries that represent basic m2m memberships
for group in qs:
ig_instance_mapping[group.name] = set(instance.hostname for instance in group.instances.all() if instance.capacity != 0)
for inst in group.instances.all():
if inst.capacity == 0:
continue
instance_ig_mapping.setdefault(inst.hostname, set())
instance_ig_mapping[inst.hostname].add(group.name)
# Get IG capacity overlap mapping
ig_ig_mapping = get_ig_ig_mapping(ig_instance_mapping, instance_ig_mapping)
return instance_ig_mapping, ig_ig_mapping
@staticmethod
def zero_out_group(graph, name, breakdown):
if name not in graph:
graph[name] = {}
graph[name]['consumed_capacity'] = 0
for capacity_type in ('execution', 'control'):
graph[name][f'consumed_{capacity_type}_capacity'] = 0
if breakdown:
graph[name]['committed_capacity'] = 0
graph[name]['running_capacity'] = 0
def capacity_values(self, qs=None, tasks=None, breakdown=False, graph=None):
"""
Returns a dictionary of capacity values for all IGs
"""
if qs is None: # Optionally BYOQS - bring your own queryset
qs = self.all().prefetch_related('instances')
instance_ig_mapping, ig_ig_mapping = self.capacity_mapping(qs=qs)
if tasks is None:
tasks = self.model.unifiedjob_set.related.related_model.objects.filter(status__in=('running', 'waiting'))
if graph is None:
graph = {group.name: {} for group in qs}
for group_name in graph:
self.zero_out_group(graph, group_name, breakdown)
for t in tasks:
# TODO: dock capacity for isolated job management tasks running in queue
impact = t.task_impact
control_groups = []
if t.controller_node:
control_groups = instance_ig_mapping.get(t.controller_node, [])
if not control_groups:
logger.warning(f"No instance group found for {t.controller_node}, capacity consumed may be innaccurate.")
if t.status == 'waiting' or (not t.execution_node and not t.is_container_group_task):
# Subtract capacity from any peer groups that share instances
if not t.instance_group:
impacted_groups = []
elif t.instance_group.name not in ig_ig_mapping:
# Waiting job in group with 0 capacity has no collateral impact
impacted_groups = [t.instance_group.name]
else:
impacted_groups = ig_ig_mapping[t.instance_group.name]
for group_name in impacted_groups:
if group_name not in graph:
self.zero_out_group(graph, group_name, breakdown)
graph[group_name]['consumed_capacity'] += impact
capacity_type = get_capacity_type(t)
graph[group_name][f'consumed_{capacity_type}_capacity'] += impact
if breakdown:
graph[group_name]['committed_capacity'] += impact
for group_name in control_groups:
if group_name not in graph:
self.zero_out_group(graph, group_name, breakdown)
graph[group_name][f'consumed_control_capacity'] += settings.AWX_CONTROL_NODE_TASK_IMPACT
if breakdown:
graph[group_name]['committed_capacity'] += settings.AWX_CONTROL_NODE_TASK_IMPACT
elif t.status == 'running':
# Subtract capacity from all groups that contain the instance
if t.execution_node not in instance_ig_mapping:
if not t.is_container_group_task:
logger.warning('Detected %s running inside lost instance, ' 'may still be waiting for reaper.', t.log_format)
if t.instance_group:
impacted_groups = [t.instance_group.name]
else:
impacted_groups = []
else:
impacted_groups = instance_ig_mapping[t.execution_node]
for group_name in impacted_groups:
if group_name not in graph:
self.zero_out_group(graph, group_name, breakdown)
graph[group_name]['consumed_capacity'] += impact
capacity_type = get_capacity_type(t)
graph[group_name][f'consumed_{capacity_type}_capacity'] += impact
if breakdown:
graph[group_name]['running_capacity'] += impact
for group_name in control_groups:
if group_name not in graph:
self.zero_out_group(graph, group_name, breakdown)
graph[group_name][f'consumed_control_capacity'] += settings.AWX_CONTROL_NODE_TASK_IMPACT
if breakdown:
graph[group_name]['running_capacity'] += settings.AWX_CONTROL_NODE_TASK_IMPACT
else:
logger.error('Programming error, %s not in ["running", "waiting"]', t.log_format)
return graph