wjddn279 commented on code in PR #58365: URL: https://github.com/apache/airflow/pull/58365#discussion_r2539104349
########## airflow-core/src/airflow/utils/gc_utils.py: ########## @@ -0,0 +1,44 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +from __future__ import annotations + +import gc +from functools import wraps + + +def with_gc_freeze(func): + """ + Freeze the GC before executing the function and unfreeze it after execution. + + This is done to prevent memory increase due to COW (Copy-on-Write) by moving all + existing objects to the permanent generation before forking the process. After the + function executes, unfreeze is called to ensure there is no impact on gc operations + in the original running process. Review Comment: > However doesn't this mean that there is no GC at all running in the task process it self? That doesn't seem like a good idea if that's the case GC runs normally, but only frozen objects are excluded from GC scanning. Specifically: - Frozen objects: Objects that existed at worker fork time (settings, config, imported libraries, etc.) → excluded from GC scanning - New objects: All objects created during worker execution → normal GC applies Frozen objects not being GC'd isn't a problem because: - They are mostly long-lived objects - They are needed for the worker's entire lifecycle - There are very few cases where they actually need to be collected And this is precisely the core reason for suggestion creating workers all at once during startup: By spawning workers as early as possible before the scheduling loop runs, we limit the scope of frozen objects to "essential initialization objects only" and prevent the scheduler's runtime state from being copied to workers. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
