The per-pod semaphore was not a cap: 3 replicas x 4 slots advertised 12
concurrent requests while the engine runs 4, and nothing coordinated the
replicas. Admission is now gated on the engine's own published gauges,
which is global by construction because every replica contends for the
same engine.
- new internal/proxy/engine_state.go: 2s scrape of the backend /metrics,
hand-rolled exposition parse (no new dependency), unhealthy after 3
consecutive failed polls, and a startup self-check that raises
observer_degraded rather than falling back to fail-open silently
- gate: park while kv_cache_usage_perc >= 0.90 or running + waiting >=
ADMISSION_MAX_INFLIGHT, bounded by ADMISSION_MAX_PARK_SECONDS, then
503 + Retry-After: 5. Unknown state fails open on purpose
- streaming requests receive SSE keep-alive heartbeats while parked
- router: extract only the top-level model field instead of unmarshalling
the whole body into a map; body cap from MAX_BODY_BYTES (1 MiB, the
model path previously inherited the 100 MB default)
- GOMAXPROCS from the cgroup CPU quota: the container is limited to 500m
while Go defaulted to the host core count and was CFS-throttled
- VLLM_MAX_QUEUE_DEPTH default 20 -> 8
- delete k8s/configmap.yaml: absent from kustomization.yaml, and the live
ConfigMap of that name carries an unrelated key
- correct the llm-client NetworkPolicy comment; llm-serving-ingress now
selects by namespace, not by that label
Observers run only for models named in ADMISSION_OBSERVE_MODELS.
reasoning is the only vLLM engine here; ornith and qwen-cpu are Ollama
and embeddings/rerank are TEI, none of which publish vllm:* series, so
observing them would hold observer_degraded at 1 permanently.
Companion change (engine flags + dashboard panels) is
riotpiao-poimen/homelab#51.
- 2.3: unknown model errors (400 + RFC 9457 problem+json with valid_models)
- 2.5: GET /v1/models endpoint (derived from config, not hardcoded)
- 2.6: POST /v1/embeddings passthrough (body-based dispatch, no rewrite)
- 2.7: POST /v1/rerank with path rewrite (/v1/rerank → /rerank)
- wire proxy.Handler in main.go (was using dummy handler)
- 140+ tests passing, race detector clean
- all requests: client → nginx → gateway → upstreams
- ready for config deployment to go live
- Dockerfile: multi-stage, distroless nonroot, CGO_ENABLED=0 static, commit
SHA stamped via VERSION build arg.
- .forgejo/workflows/ci.yaml: Forgejo reads .forgejo/, not .github/, and the
runner declares only the "docker" label. Verify job on every push; image
build and push gated to main.
- Drop .github/workflows/ci.yml — this remote is Forgejo, so it never ran.
- deployment.yaml: image from the Forgejo registry, forgejo-registry pull
secret, runAsUser 65532 to match distroless nonroot.
- kustomization.yaml: pin the tag in one place. Promoting a build is a
one-line newTag bump, never :latest.
Baseline for the Kong replacement on api.riotpiao.com. Brings the working
tree under version control for the first time: gateway source, the task
board that drives the agent runs, test fixtures, and K8s manifests.
Anchor the gateway ignore rule to the repo root. Unanchored, "gateway"
also matched the cmd/gateway/ source directory, so the program entrypoint
was excluded from every commit.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Admin Bot
rockandAdmin Bot