4 Commits

Include renames

Author	SHA1	Message	Date
poimenandrock	369d51a5c2	feat(proxy): engine-wide LLM admission control via vLLM gauges (#65) ... Closes #64 Replaces the per-pod concurrency semaphore as the effective cap with admission gated on the vLLM engine's own scheduler gauges. ## Why `internal/proxy/proxy.go:119` built a semaphore of 4 slots *inside each gateway process*, and `api-gateway` runs 3 replicas, so the advertised cap was 12 while the engine's `--max-num-seqs` is 4. Nothing coordinated replicas, and the mismatch was invisible because the queue never rejected anything (`queue_full_requests_total 0`). Measured on the engine while investigating a 10-minute TTFT: - `vllm:cache_config_info{num_gpu_blocks="9527",block_size="16"}` = **152,432 tokens** of KV capacity for the whole engine. PP=2 splits layers, so both ranks hold KV for the same tokens — not 2x that. - Uncached prefill runs at **262 tok/s** contended, 525 tok/s over 7d (`vllm:request_prefill_time_seconds_sum/count`). - One request showed `llm_ttft_seconds_sum 590.981 / count 5` at the gateway and a `(480, 960]` s entry in `vllm:request_prefill_time_seconds_bucket`, with engine-side generation throughput at 0.0-0.1 tok/s throughout: a 65k-token prompt costs ~250 s of prefill and starves every other request while it runs. - `vllm:num_preemptions_total = 0`; KV peaked at `0.5399`. Reading the engine's own gauges is global by construction, needs no lock and no new datastore, and it also accounts for callers that never pass the gateway — namespace `poimen` is admitted directly on 80/8080 by `llm-serving-ingress`, and `memory/crates/mem-cli/src/kv_warmer.rs` is such a caller. A distributed lock would have been blind to that traffic. ## Changes - **new `internal/proxy/engine_state.go`** — 2s scrape of `http://<addr>/metrics` on its own client (the proxy transports carry socket options tuned for inference), hand-rolled exposition parse, `Healthy=false` after 3 consecutive failed polls with `engine_state_stale_total` counted once at the transition, and a startup self-check that raises `observer_degraded` and logs at error level instead of degrading to fail-open silently. - **admission gate** (`internal/proxy/scheduler.go`) — parks 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 or unhealthy state fails **open** on purpose: a metrics outage must not become an inference outage. - **parked streaming heartbeat** (`internal/proxy/proxy.go`) — `stream: true` requests get 200 + `text/event-stream` and a `: keep-alive` comment every 15s while parked, so a client idle timeout does not kill a request that is legitimately waiting. Non-streaming requests park silently. - **`internal/proxy/router.go`** — extract only the top-level `model` field with a `json.Decoder` token walk instead of unmarshalling the whole body into `map[string]interface{}` (that built the entire `messages`+`tools` tree per request, ~300 KB for a 65k-token prompt). `MaxBodySize` now comes from `MAX_BODY_BYTES`; the model path previously inherited the 100 MB default at `proxy.go:93` against a 512Mi container limit. - **`cmd/gateway/main.go`** — `GOMAXPROCS` from the cgroup CPU quota (cgroup v2 `cpu.max`, v1 fallback). The container is limited to `500m` while Go defaulted to the host core count and was being CFS-throttled. - **`k8s/deployment.yaml`** — `ADMISSION_MAX_PARK_SECONDS=120`, `ADMISSION_OBSERVE_MODELS=reasoning`, `VLLM_MAX_QUEUE_DEPTH=8`, `MAX_BODY_BYTES=1048576`; corrected the `llm-client` comment (the NetworkPolicy selects by namespace now, not by that label). - **deleted `k8s/configmap.yaml`** — absent from `k8s/kustomization.yaml`, and the live ConfigMap of that name carries an unrelated key (`gw-config-fixed.yaml`) plus a stale routing table. - **`internal/observability/`** — `admission_parked_total`, `admission_rejected_total{reason}`, `engine_state_stale_total`, and gauges `vllm_requests_running` / `vllm_kv_usage_perc` (the gateway's own reading, deliberately mirroring the vLLM series so a disagreement is visible) and `observer_degraded`. Names follow the existing convention, which carries no prefix. Out of scope here, in `riotpiao-poimen/homelab#51`: the engine flags (`--max-num-partial-prefills`, `--max-long-partial-prefills`, `--long-prefill-token-threshold`, `--max-num-batched-tokens`, `--max-num-seqs`, `--gpu-memory-utilization`) and the four dashboard panels in `generate-dashboards.py`. ## Verification - `go build ./...` clean, `go vet ./...` clean. - `go test ./...` — 368 tests, 0 failures (baseline on `main` is 368/0). - `structural-check.sh` on the changed files — 0 errors, 1 pre-existing warning (`:latest` image tag). - `go.mod` / `go.sum` byte-identical: no new dependency. Not yet verified: the real-engine load test (2x65k + 15x8k against `reasoning`). That run also produces the number for `ADMISSION_MAX_PARK_SECONDS`, which today is the provisional 120; retuning it is a follow-up since it is env-only and needs no rebuild. ## Accepted risks - **No tests were added** for the observer or the gate (deliberate: production is the test bed). The mitigation is the startup self-check, which is production self-diagnosis rather than a test, plus the `observer_degraded` alert. A parse failure that outlives the first poll logs and surfaces as a metric; it does not silently disable gating without a signal. - **The cap is soft, not hard.** Between polls each replica may admit more than one request, so there is no clean bound. Exposure is low (~11,500 requests over 7d ≈ 1.4/min) and the engine remains the hard limit — a race parks requests in vLLM's queue rather than over-committing KV. - **`ADMISSION_MAX_INFLIGHT` is manually coupled to `--max-num-seqs`** and can drift silently. It defaults to `VLLM_MAX_NUM_SEQS`, and startup logs the effective value. - The first scrape waits one interval, so the first ~2s of admission is ungated (fail-open). Scraping at `Start()` instead would make the gateway contact every observed backend before any request needs one, which is wasted work and a surprising side effect. ## Rollback Revert this commit and redeploy. The admission path is fail-open, so the code can also be neutralised without reverting by unsetting `ADMISSION_OBSERVE_MODELS`. Separately, the orphaned cluster object can be removed by hand: `kubectl delete cm -n api api-gateway-config` (not a code change, deliberately left to a human). --------- Reviewed-on: #65 Co-authored-by: poimen <[email protected]>	2026-09-30 10:56:10 +00:00
rockandAdmin Bot	ef5e736673	feat(cag): CAG integration — memory adapter, warmup semaphore, cache-aware routing (#47) ... ## Summary Three issues in one PR. Implements Cache-Augmented Generation (CAG) support in the api-gw. --- ### #44 — Memory ServiceAdapter: standing-context resource - Code-registers `memory` adapter with `standing-context` resource (`POST /memory/standing-context`) - Fixes dispatcher bug: method-level `Auth.Capability` override now respected (was always using verb-derived capability) - Auth: `memory:read` required via method-level override; `memory:write` for ingest - `MEMORY_UPSTREAM_URL` env configures upstream ### #45 — Continuous batching: warmup deprioritisation + per-backend semaphore - In-memory bounded queue per backend: `VLLM_MAX_QUEUE_DEPTH` slots (default 20) - Concurrency limiter: `VLLM_MAX_NUM_SEQS` slots (default 4) - `X-Poimen-Warmup: true` → sleep `WARMUP_DELAY_MS` (default 500ms) before queue entry, strip header, add `X-Priority: low` - Queue full → 503 immediately, no goroutine pile-up - Metrics: `warmup_requests_delayed_total`, `queue_full_requests_total` at `/metrics` ### #46 — Cache-aware routing: rendezvous hash + prefix cache hit logging - `addresses: [...]` YAML field for multi-backend models (backward-compat with `address`) - `X-Poimen-Cache-Key` header → rendezvous-hash backend selection (same key = same GPU) - No cache key → round-robin - Failed backend excluded 30s; next-highest-score fallback - `vllm-num-cached-tokens` response header captured → `vllm_prefix_cache_hit_tokens_total` counter + structured log --- ## Verification ``` go build ./... ✅ go vet ./... ✅ go test -race ./... ✅ (all packages, 0 races) ``` ## Reviewer verdicts - reviewer-sre: SIGN-OFF - reviewer-quality: SIGN-OFF ## Non-blocking findings (🟡🔵 only) - `getOrCreateTransport` not goroutine-safe by design; pre-population invariant documented - `time.Sleep(warmupDelay)` holds goroutine; context-cancel alternative left as TODO - OTel trace_id not propagated into CAG log events (existing gap) Closes #44 Closes #45 Closes #46 --------- Co-authored-by: Admin Bot <[email protected]> Reviewed-on: #47	2026-09-27 11:16:40 +00:00
rockandpoimen	d439536ca9	Add TTFT & ITL Metrics for LLM Inference (#28) ... Time-to-First-Token (TTFT) and Inter-Token Latency (ITL) metrics for LLM inference observability Metrics: llm_ttft_seconds, llm_itl_seconds, llm_tokens_total Closes #31 #32 #33 --------- Co-authored-by: poimen <[email protected]> Reviewed-on: #28	2026-09-15 08:53:58 +00:00
Admin Bot	425611ec42	feat: phase 8 serviceadapter crd rollout (32/33 tasks)	2026-08-26 13:47:36 -07:00