fix: align cpu readiness client timeout #140

Merged
m merged 1 commits from codex/per-provider-readiness-timeouts into main 2026-08-22 18:03:16 +00:00
3 changed files with 27 additions and 3 deletions
+2 -1
View File
@@ -85,6 +85,7 @@ func main() {
gateway := cognitiveGateway{resolver: resolver, providers: providers, builtinLocal: builtinLocal, externalRecovery: externalRecovery, now: time.Now, allowCanary: os.Getenv("COGNITIVE_ALLOW_CANARY") == "true", canaryMaxOutputBytes: 4096, admission: newHTTPAdmissionClient(envOr("CORE_RUNTIME_URL", "http://core-runtime:8080"), &http.Client{Timeout: 5 * time.Second})}
mux.Handle("/healthz", health.Handler{Runtime: "cognitive-worker", SemanticGeneration: 1, ArchitectureGeneration: 1})
readinessClient := &http.Client{Timeout: 2 * time.Second}
cpuReadinessClient := &http.Client{Timeout: 10 * time.Second}
lifeboatURL := strings.TrimSuffix(strings.TrimRight(envOr("LOCAL_LIFEBOAT_PROVIDER_URL", "http://cognitive-lifeboat:8086/v1"), "/"), "/v1") + "/health"
fleetClient := &http.Client{Timeout: 2 * time.Second}
preferredFleet := capability.FleetTier(envOr("COGNITIVE_PREFERRED_FLEET_TIER", string(capability.FleetTierHeavy)))
@@ -93,7 +94,7 @@ func main() {
cpuFleetHeaders["Authorization"] = "Bearer " + cpuAPIKey
}
cpuEndpoint := strings.TrimRight(envOr("LOCAL_CPU_PROVIDER_URL", "http://docker.cin.su:18082/v1"), "/") + "/models"
cpuHealthProbe := health.OptionalHTTPProbeWithHeaders(readinessClient, cpuEndpoint, cpuFleetHeaders)
cpuHealthProbe := health.OptionalHTTPProbeWithHeaders(cpuReadinessClient, cpuEndpoint, cpuFleetHeaders)
cpuHealthProbe.Timeout = 10 * time.Second
readinessProbes := []health.Probe{
health.HTTPProbe(readinessClient, strings.TrimRight(envOr("CORE_RUNTIME_URL", "http://core-runtime:8080"), "/")+"/healthz"),
+2 -2
View File
@@ -1,5 +1,5 @@
{
"task_file": "dev/tasks/DEV-000166-per-provider-readiness-timeouts.json",
"task_id": "DEV-000166",
"task_file": "dev/tasks/DEV-000167-cpu-readiness-client-timeout.json",
"task_id": "DEV-000167",
"task_revision": 1
}
@@ -0,0 +1,23 @@
{
"task_id": "DEV-000167",
"task_revision": 1,
"architecture_revision": "ADR-0018",
"base_repo_revision": "c7b57daaf8233b0f0e79c77c02e92040378921d5",
"risk_tier": "E",
"intent": "Use the bounded CPU provider HTTP client for the authenticated readiness probe so /readyz reports the same truthful result as fleet readiness.",
"requirement_ids": ["WP-601", "Model fleet/failover"],
"allowed_change_scope": {"modules": ["cognitive-worker"], "paths": ["cmd/cognitive-worker/main.go", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000167-cpu-readiness-client-timeout.json"]},
"forbidden_change_scope": {"paths": ["architecture/**", "security/**", "database/**", "deploy/**", "contracts/**", "adapters/**", ".gitea/**"]},
"allowed_architecture_delta": {"new_cross_module_dependencies": 0, "new_canonical_owners": 0, "new_privileges": 0, "new_third_party_dependencies": 0, "new_database_owner_roles": 0},
"contracts": {"consume": ["existing readiness Probe contract", "existing authenticated CPU fleet endpoint"], "may_change": []},
"semantic_owner": "cognitive-worker",
"canonical_state": [],
"state_transitions": ["CPU provider response -> consistent /readyz and fleet readiness state"],
"side_effects_and_privileges": ["no new privilege; existing provider auth only"],
"persistence_and_migrations": ["none"],
"failure_retry_restart": ["CPU readiness remains bounded at 10 seconds; opportunistic GPU remains 2 seconds"],
"assumptions": ["The CPU provider is authenticated with the existing file-backed API key"],
"acceptance_scenarios": ["/readyz can report CPU AVAILABLE when fleet readiness reports CPU AVAILABLE", "GPU remains bounded", "Lifeboat fallback remains unchanged"],
"required_tests": ["go test ./...", "go vet ./...", "preflight", "check-diff", "check-architecture", "independent", "git diff --check", "CI", "production /readyz and fleet readiness"],
"terminal_gap_states": ["BLOCKED_BY_MISSING_CAPABILITY", "BLOCKED_BY_CONFLICT"]
}