merge cognitive readiness budget fix

cmd/core-runtime/main.go

@@ -154,13 +154,14 @@ func main() {
 	}
 	mux.Handle("/healthz", health.Handler{Runtime: "core-runtime", SemanticGeneration: 1, ArchitectureGeneration: 1})
 	readinessClient := &http.Client{Timeout: 2 * time.Second}
+	cognitiveReadinessClient := &http.Client{Timeout: 20 * time.Second}
 	mux.Handle("/readyz", health.ReadinessHandler{Runtime: "core-runtime", Probes: []health.Probe{
 		health.TCPProbe("postgres", envOr("POSTGRES_HOST", "postgres"), envOr("POSTGRES_PORT", "5432")),
 		health.HTTPProbe(readinessClient, strings.TrimRight(envOr("CONTINUITY_RUNTIME_URL", "http://continuity-runtime:8081"), "/")+"/healthz"),
 		// A running worker is not necessarily ready to serve cognitive traffic.
 		// Keep deterministic paths alive, but do not advertise overall readiness
 		// while the cognitive dependency is still warming.
-		health.HTTPProbe(readinessClient, strings.TrimRight(envOr("COGNITIVE_RUNTIME_URL", "http://cognitive-worker:8083"), "/")+"/readyz"),
+		health.HTTPProbe(cognitiveReadinessClient, strings.TrimRight(envOr("COGNITIVE_RUNTIME_URL", "http://cognitive-worker:8083"), "/")+"/readyz"),
 	}})
 	admissionPolicy := capability.AdmissionPolicy{ID: envOr("COGNITIVE_ADMISSION_POLICY_ID", "creator-advisory-local-v1"), MaxConcurrent: envInt("COGNITIVE_ADMISSION_MAX_CONCURRENT", 1), MaxAcceptedPerHour: envInt("COGNITIVE_ADMISSION_MAX_ACCEPTED_PER_HOUR", 60), MaxInputBytes: envInt("COGNITIVE_ADMISSION_MAX_INPUT_BYTES", 80<<10), MaxOutputBytes: envInt("COGNITIVE_ADMISSION_MAX_OUTPUT_BYTES", 16<<10)}
 	mux.HandleFunc("POST /v1/internal/cognitive/admission", func(w http.ResponseWriter, r *http.Request) {

deploy/docker-compose.prod.yml

@@ -267,7 +267,7 @@ services:
     healthcheck:
       test: ["CMD-SHELL", "wget -qO- http://localhost:8083/healthz >/dev/null && wget -qO- http://localhost:8083/readyz >/dev/null"]
       interval: 10s
-      timeout: 5s
+      timeout: 20s
       retries: 20
 
   cognitive-lifeboat: