Update dynamic-lora-sidecar to expose metrics to track loaded adapters

[shotarok]

Resolves #600

Update dynamic-lora-sidecar to expose metrics to track loaded adapters. lora_syncer_adapter_status is a binary metric with the adapter_name label. I confirmed sidecar.py ran an HTTP server with the port 8080 that returned metrics.

curl http://localhost:8080/metrics

• Run sidecar.py

(venv) laborant@dev-machine:sidecar$ pwd
/home/laborant/gateway-api-inference-extension/tools/dynamic-lora-sidecar/sidecar
(venv) laborant@dev-machine:sidecar$ uv run --with-requirements ../requirements.txt python sidecar.py --config ../configmap.yaml
2025-06-13 04:59:09 - INFO - sidecar.py:118 -  Settings initialized: health check timeout=300s, interval=2s, reconcile trigger=5s
2025-06-13 04:59:09 - INFO - sidecar.py:332 -  Starting metrics server on port 8080
2025-06-13 04:59:09 - INFO - sidecar.py:335 -  Running initial reconcile for config map ../configmap.yaml
2025-06-13 04:59:09 - INFO - sidecar.py:281 -  reconciling model server localhost:8000 with config stored at ../configmap.yaml

• Run curl

laborant@dev-machine:~$ curl http://localhost:8080/metrics
# HELP python_gc_objects_collected_total Objects collected during gc
# TYPE python_gc_objects_collected_total counter
python_gc_objects_collected_total{generation="0"} 1473.0
python_gc_objects_collected_total{generation="1"} 168.0
python_gc_objects_collected_total{generation="2"} 0.0
# HELP python_gc_objects_uncollectable_total Uncollectable objects found during GC
# TYPE python_gc_objects_uncollectable_total counter
python_gc_objects_uncollectable_total{generation="0"} 0.0
python_gc_objects_uncollectable_total{generation="1"} 0.0
python_gc_objects_uncollectable_total{generation="2"} 0.0
# HELP python_gc_collections_total Number of times this generation was collected
# TYPE python_gc_collections_total counter
python_gc_collections_total{generation="0"} 76.0
python_gc_collections_total{generation="1"} 6.0
python_gc_collections_total{generation="2"} 0.0
# HELP python_info Python platform information
# TYPE python_info gauge
python_info{implementation="CPython",major="3",minor="12",patchlevel="3",version="3.12.3"} 1.0
# HELP process_virtual_memory_bytes Virtual memory size in bytes.
# TYPE process_virtual_memory_bytes gauge
process_virtual_memory_bytes 1.99852032e+08
# HELP process_resident_memory_bytes Resident memory size in bytes.
# TYPE process_resident_memory_bytes gauge
process_resident_memory_bytes 4.1213952e+07
# HELP process_start_time_seconds Start time of the process since unix epoch in seconds.
# TYPE process_start_time_seconds gauge
process_start_time_seconds 1.74979074849e+09
# HELP process_cpu_seconds_total Total user and system CPU time spent in seconds.
# TYPE process_cpu_seconds_total counter
process_cpu_seconds_total 0.5700000000000001
# HELP process_open_fds Number of open file descriptors.
# TYPE process_open_fds gauge
process_open_fds 9.0
# HELP process_max_fds Maximum number of open file descriptors.
# TYPE process_max_fds gauge
process_max_fds 1024.0
# HELP lora_syncer_adapter_status Status of LoRA adapters (1=loaded, 0=not_loaded)
# TYPE lora_syncer_adapter_status gauge

pytest

(venv) laborant@dev-machine:sidecar$ uv run --with-requirements ../requirements.txt --with pytest pytest .
==================================================================== test session starts =====================================================================
platform linux -- Python 3.12.3, pytest-8.4.0, pluggy-1.6.0
rootdir: /home/laborant/gateway-api-inference-extension/tools/dynamic-lora-sidecar/sidecar
plugins: anyio-4.9.0
collected 6 items                                                                                                                                            

test_sidecar.py ......                                                                                                                                 [100%]

===================================================================== 6 passed in 0.26s ======================================================================

[netlify]

✅ Deploy Preview for gateway-api-inference-extension ready!

Name	Link
🔨 Latest commit	7269f88
🔍 Latest deploy log	https://app.netlify.com/projects/gateway-api-inference-extension/deploys/68509a0446018a0008443023
😎 Deploy Preview	https://deploy-preview-980--gateway-api-inference-extension.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[k8s-ci-robot]

Hi @shotarok. Thanks for your PR.

I'm waiting for a kubernetes-sigs member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[danehans]

@shotarok thanks for the PR. If the lora syncer metrics were exposed on port 9090, could they be remotely scraped by following the metrics guide? Either way, this guide should be updated with a new LoRA Syncer section that provides details similar to the existing EPP metrics.

As a follow on, e2e tests should be updated to include test cases for scraping EPP and lora syncer metrics with assertions for the expected metrics. Feel free to create an issue and assign yourself if you can help with this.

[shotarok]

@danehans Thank you for your feedback! If we use the same port in both main and sidecar containers, I think scraping of the /metrics endpoint won't work consistently. I'll update the docs in this PR to add the LoRA Syncer section.

As for the e2e tests, until the vLLM simulator supports the following endpoints to load/unload a LoRA adapter, we can't write an e2e test for LoRA Syncer metrics with the vLLM simulator. I think we can update an e2e test for EPP even now, so I'll create separate issues as a follow-up.

• /v1/unload_lora_adapter
• /v1/load_lora_adapter

[shotarok]

@danehans I updated the documentation about the metrics and created two issues for the e2e tests. I found hermetic_test for EPP's metrics. If the e2e test feels unnecessary for EPP's metrics, please feel free to close that issue. Thanks!

[danehans]

xref: llm-d/llm-d-inference-sim#58 for lora api endpoint support.

[danehans]

A few nits that are non-blocking. @JeffLuoo PTAL since you have experience with GIE metrics.

[danehans]

/lgtm

[kfswain]

/ok-to-test

[JeffLuoo]

/lgtm

[danehans]

/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: danehans, shotarok

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [danehans]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment