[router] Update metrics when request completes

[ZhangShuaiyi]

Motivation

Updates sgl_router_running_requests when request completes.

sglang.bench_serving:

python3 -m sglang.bench_serving --backend sglang --num-prompt 20 --dataset-name random-ids --random-input 512 --random-output 512

Behavior before this change:
The metric sgl_router_running_requests was only updated in select_generate_worker_from_text. When sglang.bench_serving finishes, the router logs show no running queues

2025-07-09 19:27:39  INFO sglang_router_rs::tree: src/tree.rs:430: Before eviction - Used size per tenant:
2025-07-09 19:27:39  INFO sglang_router_rs::tree: src/tree.rs:432: Tenant: http://127.0.0.1:30002, Size: 11648
2025-07-09 19:27:39  INFO sglang_router_rs::tree: src/tree.rs:432: Tenant: http://127.0.0.1:30001, Size: 9819
2025-07-09 19:27:39  INFO sglang_router_rs::tree: src/tree.rs:481: After eviction - Used size per tenant:
2025-07-09 19:27:39  INFO sglang_router_rs::tree: src/tree.rs:483: Tenant: http://127.0.0.1:30002, Size: 11648
2025-07-09 19:27:39  INFO sglang_router_rs::tree: src/tree.rs:483: Tenant: http://127.0.0.1:30001, Size: 9819
2025-07-09 19:27:39  INFO sglang_router_rs::router: src/router.rs:252: Processed Queue: {"http://127.0.0.1:30002": 12, "http://127.0.0.1:30001": 9}
2025-07-09 19:27:39  INFO sglang_router_rs::router: src/router.rs:256: Running Queue: {"http://127.0.0.1:30002": 0, "http://127.0.0.1:30001": 0}

But sgl_router_running_requests in the metrics remains non-zero.

(venvsglang) root@nodedev:~# for i in {1..100};do curl -s 127.0.0.1:29000/metrics | grep running;sleep 1;done
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 1
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 1
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 9
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 11
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 11
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 9
...
...
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 11
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 9

Behavior after this change:
The sgl_router_running_requests metric now correctly drops to zero.

(venvsglang) root@nodedev:~# for i in {1..100};do curl -s 127.0.0.1:29000/metrics | grep running;sleep 1;done
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 1
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 0
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 10
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 9
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 9
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 10
...
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 2
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 4
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 2
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 3
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 2
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 2
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 0
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 1
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 0
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 0
# TYPE sgl_router_running_requests gauge
sgl_router_running_requests{worker="http://127.0.0.1:30001"} 0
sgl_router_running_requests{worker="http://127.0.0.1:30002"} 0

Modifications

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[gemini-code-assist]

Summary of Changes

Hello @ZhangShuaiyi, 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 focuses on improving the accuracy of the router's operational metrics. Specifically, it resolves a bug where the sgl_router_running_requests metric would incorrectly remain non-zero even after all active requests had finished, providing a more reliable real-time representation of the router's workload.

Highlights

• Metric Accuracy Fix: I've addressed an issue where the sgl_router_running_requests gauge metric was not being decremented when requests completed. This fix ensures the metric accurately reflects the true number of running requests by updating it upon request completion for both standard and streaming requests.

Using Gemini Code Assist

The 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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

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 .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist is currently in preview and 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 to provide feedback.

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

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request correctly fixes a bug where the sgl_router_running_requests metric was not updated when requests complete. The change adds the necessary metric updates for both streaming and non-streaming requests.

My review focuses on a code quality improvement. The logic for updating the metric is duplicated in two places. I've provided a suggestion to refactor this into a shared helper to improve maintainability and prevent potential future inconsistencies. Well done on fixing this bug!