[Doc] Add doc for LMCache MP mode operator#2731
Conversation
Signed-off-by: royyhuang <roy.y.huang@gmail.com>
Signed-off-by: royyhuang <roy.y.huang@gmail.com>
Summary of ChangesHello, 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 introduces extensive documentation for the LMCache Kubernetes operator, designed to streamline the deployment and management of LMCache multiprocess servers. The new guide explains how the operator simplifies complex Kubernetes configurations through a custom resource, offering benefits such as automated resource provisioning, simplified service discovery, and robust validation. It serves as a complete resource for users looking to leverage the operator for efficient LMCache deployments. Highlights
🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console. Changelog
Activity
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
|
There was a problem hiding this comment.
Code Review
This pull request introduces excellent and comprehensive documentation for the LMCache Kubernetes operator. The new documentation is well-structured, covering everything from installation and configuration to production best practices. My review includes a few suggestions to improve clarity and promote best practices, particularly regarding production configurations for container images and Prometheus monitoring.
| The operator defaults the container image to ``lmcache/vllm-openai:latest``. | ||
| Override with ``spec.image.repository`` and ``spec.image.tag`` to pin a | ||
| specific version. |
There was a problem hiding this comment.
The note explains that the default image can be overridden but doesn't clarify why a user should consider changing the repository. The default lmcache/vllm-openai image is large as it contains both vLLM and LMCache. For LMCache server pods in production, a smaller, dedicated image like lmcache/standalone is more appropriate. Please expand this note to provide this guidance for production deployments.
| The operator defaults the container image to ``lmcache/vllm-openai:latest``. | |
| Override with ``spec.image.repository`` and ``spec.image.tag`` to pin a | |
| specific version. | |
| The operator defaults the container image to ``lmcache/vllm-openai:latest``. | |
| For production, consider overriding ``spec.image.repository`` to a dedicated | |
| image like ``lmcache/standalone`` and pin a specific ``spec.image.tag``. |
| hostIPC: true | ||
| containers: | ||
| - name: vllm | ||
| image: lmcache/vllm-openai:latest |
There was a problem hiding this comment.
The example for connecting vLLM uses the :latest image tag, which is not recommended for production environments as it can lead to unexpected behavior when the image is updated. To promote best practices and ensure reproducibility, it's better to use a specific version tag. The production example later in this document uses v0.1.0, which would be a good choice here as well.
| image: lmcache/vllm-openai:latest | |
| image: lmcache/vllm-openai:v0.1.0 |
| podAnnotations: | ||
| prometheus.io/scrape: "true" | ||
| prometheus.io/port: "9090" |
There was a problem hiding this comment.
The production example enables the ServiceMonitor and also adds prometheus.io/scrape pod annotations. This is redundant. When using the Prometheus Operator, the ServiceMonitor is the idiomatic way to configure scraping, and it targets the metrics service created by the LMCache operator. The pod annotations are for a different discovery mechanism (Prometheus's native Kubernetes SD). Including both can be confusing. To simplify and align with best practices for the Prometheus Operator, I recommend removing the podAnnotations.
* [doc] add mp operator documentation * [doc] add operator to the index after merge Signed-off-by: royyhuang <roy.y.huang@gmail.com>
* [doc] add mp operator documentation * [doc] add operator to the index after merge Signed-off-by: royyhuang <roy.y.huang@gmail.com> Signed-off-by: Aaron Wu <aaron.wu@dell.com>
* [doc] add mp operator documentation * [doc] add operator to the index after merge Signed-off-by: royyhuang <roy.y.huang@gmail.com>
* [doc] add mp operator documentation * [doc] add operator to the index after merge Signed-off-by: royyhuang <roy.y.huang@gmail.com>
What this PR does / why we need it:
Adding documentation on how to use LMCache K8s operator with MP mode.
Special notes for your reviewers:
If applicable: