Remove Gunicorn and use Uvicorn only for gateway #530
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
yunfeng-scale
changed the title
yunfeng-scale
changed the title
edgan8
reviewed
May 31, 2024
| "--workers", | ||
| f"{num_workers}", | ||
| "1", # Let the Kubernetes deployment handle the number of pods |
There was a problem hiding this comment.
Will this reduce the amount of traffic we can receive per-pod? Why not keep it at 4?
There was a problem hiding this comment.
this is to remove load balancing within pod
There was a problem hiding this comment.
are we increasing the number of pods by 4 to compensate?
There was a problem hiding this comment.
that's the initial plan
There was a problem hiding this comment.
tbh i think llm engine is overprovisioned
In addition to:
Let's also monitor post-rollout 👀 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Pull Request Summary
in kubernetes environment we don't really need multiple workers in the same pod, rather it's simpler to just have kubernetes autoscale the number of pods. based on some internal benchmarks gunicorn has some known load balancing issues, also removing this layer results in less error and better latency
Test Plan and Usage Guide
will run simple load testing for get requests with and without gunicorn