[Pools] Add heterogeneity support. #8315
Conversation
Summary of ChangesHello @lloyd-brown, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly enhances the 'Pools' feature by introducing full support for heterogeneous resource configurations. This allows users to define jobs with flexible resource requirements (e.g., 'any_of' T4 or A100 GPUs), and the system will intelligently select and record the specific resource used from the available pool workers. The changes involve core logic updates for resource handling, database integration for tracking resolved resources, and the removal of prior limitations on heterogeneous pools, all validated by new, dedicated smoke tests. Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
|
/smoke-test -k test_pools_heterogeneous |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces heterogeneity support for resource pools, a significant enhancement. The changes are well-structured and include updates to resource configuration handling, scheduling logic for heterogeneous resources, and database state management. The new tests for heterogeneity and resource-aware scheduling are comprehensive. However, I've identified a critical issue in the _task_fits function, which is now used by the new scheduling logic. This function incorrectly compares CPU and memory requirements, which could lead to incorrect scheduling decisions.
| if free_resources_on_worker is not None: | ||
| # Find which task resource fits on this worker | ||
| for task_res in task_resources_list: | ||
| if _task_fits(task_res, free_resources_on_worker): |
There was a problem hiding this comment.
The _task_fits function, which is called here, appears to have a bug in how it compares CPU and memory resources. It compares them as strings, which can lead to incorrect results. For example, in Python, '10' < '2' is True, so a task requiring 10 CPUs could be incorrectly scheduled on a worker with only 2 free CPUs. This could cause scheduling failures or resource contention.
The cpus and memory values from Resources objects should be parsed into numbers before comparison. The + suffix for minimum requirements should also be handled during parsing.
|
/smoke-test -k test_pools_heterogeneous |
|
/smoke-test -k test_pools |
Tested (run the relevant ones):
bash format.sh/smoke-test(CI) orpytest tests/test_smoke.py(local)/smoke-test -k test_name(CI) orpytest tests/test_smoke.py::test_name(local)/quicktest-core(CI) orpytest tests/smoke_tests/test_backward_compat.py(local)