ed7ad49d3a
P3.2 — Multi-node deployment: - Added node-worker service to docker-compose.node1.yml (NODE_ID=noda1) - NCS NODA1 now has NODE_WORKER_URL for metrics collection - Fixed NODE_ID consistency: router NODA1 uses 'noda1' - NODA2 node-worker/NCS gets NCS_REPORT_URL for latency reporting P3.3 — NATS accounts/auth (opt-in config): - config/nats-server.conf with 3 accounts: SYS, FABRIC, APP - Per-user topic permissions (router, ncs, node_worker) - Leafnode listener :7422 with auth - Not yet activated (requires credential provisioning) P3.4 — Prometheus counters: - Router /fabric_metrics: caps_refresh, caps_stale, model_select, offload_total, breaker_state, score_ms histogram - Node Worker /prom_metrics: jobs_total, inflight gauge, latency_ms histogram - NCS /prom_metrics: runtime_health, runtime_p50/p95, node_wait_ms - All bound to 127.0.0.1 (not externally exposed) Made-with: Cursor
83 lines
2.4 KiB
Python
83 lines
2.4 KiB
Python
"""Prometheus metrics for Fabric routing layer.
|
|
|
|
Exposed via /fabric_metrics (separate from main /metrics to avoid conflicts).
|
|
Falls back to no-op counters if prometheus_client is not installed.
|
|
"""
|
|
import logging
|
|
import time
|
|
from typing import Optional
|
|
|
|
logger = logging.getLogger("fabric_metrics")
|
|
|
|
try:
|
|
from prometheus_client import Counter, Gauge, Histogram, CollectorRegistry, generate_latest
|
|
PROM_AVAILABLE = True
|
|
REGISTRY = CollectorRegistry()
|
|
|
|
caps_refresh = Counter(
|
|
"fabric_caps_refresh_total", "Capabilities refresh attempts",
|
|
["status"], registry=REGISTRY,
|
|
)
|
|
caps_stale = Counter(
|
|
"fabric_caps_stale_total", "Stale capabilities used",
|
|
registry=REGISTRY,
|
|
)
|
|
model_select = Counter(
|
|
"fabric_model_select_total", "Model selection decisions",
|
|
["chosen_node", "chosen_runtime", "type"], registry=REGISTRY,
|
|
)
|
|
offload_total = Counter(
|
|
"fabric_offload_total", "Offload attempts",
|
|
["status", "node", "type"], registry=REGISTRY,
|
|
)
|
|
breaker_state = Gauge(
|
|
"fabric_breaker_state", "Circuit breaker state (1=open)",
|
|
["node", "type"], registry=REGISTRY,
|
|
)
|
|
score_hist = Histogram(
|
|
"fabric_score_ms", "Model selection score distribution",
|
|
buckets=[100, 250, 500, 1000, 2000, 5000, 10000],
|
|
registry=REGISTRY,
|
|
)
|
|
|
|
except ImportError:
|
|
PROM_AVAILABLE = False
|
|
REGISTRY = None
|
|
logger.info("prometheus_client not installed, fabric metrics disabled")
|
|
|
|
|
|
def inc_caps_refresh(status: str):
|
|
if PROM_AVAILABLE:
|
|
caps_refresh.labels(status=status).inc()
|
|
|
|
|
|
def inc_caps_stale():
|
|
if PROM_AVAILABLE:
|
|
caps_stale.inc()
|
|
|
|
|
|
def inc_model_select(node: str, runtime: str, req_type: str):
|
|
if PROM_AVAILABLE:
|
|
model_select.labels(chosen_node=node, chosen_runtime=runtime, type=req_type).inc()
|
|
|
|
|
|
def inc_offload(status: str, node: str, req_type: str):
|
|
if PROM_AVAILABLE:
|
|
offload_total.labels(status=status, node=node, type=req_type).inc()
|
|
|
|
|
|
def set_breaker(node: str, req_type: str, is_open: bool):
|
|
if PROM_AVAILABLE:
|
|
breaker_state.labels(node=node, type=req_type).set(1 if is_open else 0)
|
|
|
|
|
|
def observe_score(score: int):
|
|
if PROM_AVAILABLE:
|
|
score_hist.observe(score)
|
|
|
|
|
|
def get_metrics_text() -> Optional[bytes]:
|
|
if PROM_AVAILABLE and REGISTRY:
|
|
return generate_latest(REGISTRY)
|
|
return None
|