make orchestrator run as long as any tasks are pending

This commit is contained in:
Nick Sweeting 2024-11-16 06:42:04 -08:00
parent a4635fe65d
commit 210fd935d7
No known key found for this signature in database
5 changed files with 33 additions and 15 deletions

View file

@ -798,7 +798,7 @@ ArchiveBox bundles industry-standard tools like [Google Chrome](https://github.c
<li>Database: <a href="https://docs.djangoproject.com/en/5.1/ref/databases/#sqlite-notes">Django ORM</a> saving to <a href="https://www.sqlite.org/mostdeployed.html">SQLite3</a> <code>./data/index.sqlite</code></li>
<li>Job Queue: <a href="https://huey.readthedocs.io/">Huey</a> using <code>./data/queue.sqlite3</code> under <code>supervisord</code></li>
<li>Build/test/lint: <a href="https://github.com/pdm-project/pdm"><code>pdm</code></a> / <code>mypy</code>+<code>pyright</code>+<code>pytest</code> / <code>ruff</code></li>
<li>Subdependencies: <a href="https://github.com/ArchiveBox/pydantic-pkgr"><code>pydantic-pkgr</code></a> installs apt/brew/pip/npm pkgs at runtime (e.g. <code>yt-dlp</code>, <code>singlefile</code>, <code>readability</code>, <code>git</code>)</li>
<li>Subdependencies: <a href="https://github.com/ArchiveBox/abx-pkg"><code>abx-pkg</code></a> installs apt/brew/pip/npm pkgs at runtime (e.g. <code>yt-dlp</code>, <code>singlefile</code>, <code>readability</code>, <code>git</code>)</li>
</ul>

View file

@ -55,15 +55,15 @@ load_vendored_pkgs()
import abx # noqa
import abx_spec_archivebox # noqa
import abx_spec_config # noqa
import abx_spec_pydantic_pkgr # noqa
import abx_spec_abx_pkg # noqa
import abx_spec_django # noqa
import abx_spec_searchbackend # noqa
abx.pm.add_hookspecs(abx_spec_config.PLUGIN_SPEC)
abx.pm.register(abx_spec_config.PLUGIN_SPEC())
abx.pm.add_hookspecs(abx_spec_pydantic_pkgr.PLUGIN_SPEC)
abx.pm.register(abx_spec_pydantic_pkgr.PLUGIN_SPEC())
abx.pm.add_hookspecs(abx_spec_abx_pkg.PLUGIN_SPEC)
abx.pm.register(abx_spec_abx_pkg.PLUGIN_SPEC())
abx.pm.add_hookspecs(abx_spec_django.PLUGIN_SPEC)
abx.pm.register(abx_spec_django.PLUGIN_SPEC())

View file

@ -0,0 +1,2 @@
__package__ = 'archivebox.actors'
__order__ = 100

View file

@ -465,7 +465,7 @@ class ActorType(Generic[ModelType]):
def tick(self, obj_to_process: ModelType) -> None:
"""Call the object.sm.tick() method to process the object"""
print(f'[blue]🏃‍♂️ {self}.tick()[/blue] {obj_to_process}')
print(f'[blue]🏃‍♂️ {self}.tick()[/blue] {obj_to_process.ABID} {obj_to_process.status} {obj_to_process.retry_at}')
# get the StateMachine instance from the object
obj_statemachine = self._get_state_machine_instance(obj_to_process)
@ -500,17 +500,22 @@ class ActorType(Generic[ModelType]):
# abx.pm.hook.on_actor_shutdown(actor=self, last_obj=last_obj, last_error=last_error)
def on_tick_start(self, obj_to_process: ModelType) -> None:
print(f'🏃‍♂️ {self}.on_tick_start() {obj_to_process}')
print(f'🏃‍♂️ {self}.on_tick_start() {obj_to_process.ABID} {obj_to_process.status} {obj_to_process.retry_at}')
# abx.pm.hook.on_actor_tick_start(actor=self, obj_to_process=obj)
# self.timer = TimedProgress(self.MAX_TICK_TIME, prefix=' ')
pass
def on_tick_end(self, obj_to_process: ModelType) -> None:
print(f'🏃‍♂️ {self}.on_tick_end() {obj_to_process}')
print(f'🏃‍♂️ {self}.on_tick_end() {obj_to_process.ABID} {obj_to_process.status} {obj_to_process.retry_at}')
# abx.pm.hook.on_actor_tick_end(actor=self, obj_to_process=obj_to_process)
# self.timer.end()
pass
# import ipdb; ipdb.set_trace()
def on_tick_exception(self, obj_to_process: ModelType, error: Exception) -> None:
print(f'[red]🏃‍♂️ {self}.on_tick_exception()[/red] {obj_to_process}: [red]{type(error).__name__}: {error}[/red]')
print(f'[red]🏃‍♂️ {self}.on_tick_exception()[/red] {obj_to_process.ABID} {obj_to_process.status} {obj_to_process.retry_at}: [red]{type(error).__name__}: {error}[/red]')
# abx.pm.hook.on_actor_tick_exception(actor=self, obj_to_process=obj_to_process, error=error)

View file

@ -97,6 +97,15 @@ class Orchestrator:
orphaned_objects.extend(model.objects.filter(retry_at__lt=timezone.now()).exclude(id__in=all_queued_ids))
return orphaned_objects
@classmethod
def has_future_objects(cls, all_queues) -> bool:
# returns a list of objects that are in the queues of all actor types but not in the queues of any other actor types
return any(
queue.filter(retry_at__gt=timezone.now()).exists()
for queue in all_queues.values()
)
def on_startup(self):
if self.mode == 'thread':
self.pid = get_native_id()
@ -111,8 +120,8 @@ class Orchestrator:
# abx.pm.hook.on_orchestrator_shutdown(self)
def on_tick_started(self, all_queues):
# total_pending = sum(queue.count() for queue in all_queues.values())
# print(f'👨‍✈️ {self}.on_tick_started()', f'total_pending={total_pending}')
total_pending = sum(queue.count() for queue in all_queues.values())
print(f'👨‍✈️ {self}.on_tick_started()', f'total_pending={total_pending}')
# abx.pm.hook.on_orchestrator_tick_started(self, actor_types, all_queues)
pass
@ -123,15 +132,15 @@ class Orchestrator:
# abx.pm.hook.on_orchestrator_tick_finished(self, actor_types, all_queues)
def on_idle(self, all_queues):
# print(f'👨‍✈️ {self}.on_idle()')
print(f'👨‍✈️ {self}.on_idle()', f'idle_count={self.idle_count}')
# abx.pm.hook.on_orchestrator_idle(self)
# check for orphaned objects left behind
if self.idle_count == 60:
orphaned_objects = self.get_orphaned_objects(all_queues)
if orphaned_objects:
print('[red]👨‍✈️ WARNING: some objects may not be processed, no actor has claimed them after 60s:[/red]', orphaned_objects)
if self.idle_count > 5 and self.exit_on_idle:
raise KeyboardInterrupt('Orchestrator has no more tasks to process, exiting')
print('[red]👨‍✈️ WARNING: some objects may not be processed, no actor has claimed them after 30s:[/red]', orphaned_objects)
if self.idle_count > 3 and self.exit_on_idle and not self.has_future_objects(all_queues):
raise KeyboardInterrupt('✅ All tasks completed, exiting')
def runloop(self):
from archivebox.config.django import setup_django
@ -153,6 +162,8 @@ class Orchestrator:
all_spawned_actors = []
for actor_type, queue in all_queues.items():
next_obj = queue.first()
print(f'🏃‍♂️ {self}.runloop() {actor_type.__name__.ljust(20)} queue={str(queue.count()).ljust(3)} next={next_obj.abid if next_obj else "None"} {next_obj.status if next_obj else "None"} {(timezone.now() - next_obj.retry_at).total_seconds() if next_obj else "None"}')
try:
existing_actors = actor_type.get_running_actors()
all_existing_actors.extend(existing_actors)
@ -168,7 +179,7 @@ class Orchestrator:
if not any(queue.exists() for queue in all_queues.values()):
self.on_idle(all_queues)
self.idle_count += 1
time.sleep(1)
time.sleep(0.5)
else:
self.idle_count = 0