Compare commits

...
Author SHA1 Message Date
m ae06d33ccc fix: align cpu readiness client timeout
CI / test (push) Successful in 34s
CI / test (pull_request) Successful in 33s
2026-08-22 21:00:46 +03:00
m c7b57daaf8 Merge pull request 'fix: use provider-specific readiness timeouts' (#139) from codex/per-provider-readiness-timeouts into main
CI / test (push) Successful in 38s
2026-08-22 17:57:27 +00:00
m 8aa0123047 chore: format readiness health server
CI / test (push) Successful in 31s
CI / test (pull_request) Successful in 35s
2026-08-22 20:55:48 +03:00
m 4a3d348e39 fix: use provider-specific readiness timeouts
CI / test (push) Failing after 25s
CI / test (pull_request) Failing after 23s
2026-08-22 20:52:40 +03:00
m 4086b75224 Merge pull request 'fix: authenticate cognitive readiness probe' (#138) from codex/authorized-cpu-health-probe into main
CI / test (push) Successful in 43s
2026-08-22 17:47:25 +00:00
m 4f89c46545 fix: authenticate cognitive readiness probe
CI / test (push) Successful in 36s
CI / test (pull_request) Successful in 34s
2026-08-22 20:46:03 +03:00
m 5d5afc8888 Merge pull request 'fix: authenticate fleet readiness probe' (#137) from codex/authorized-fleet-readiness into main
CI / test (push) Successful in 48s
2026-08-22 17:41:55 +00:00
8 changed files with 134 additions and 11 deletions
+6 -1
View File
@@ -18,6 +18,7 @@ type fleetEndpoint struct {
Tier capability.FleetTier
URL string
Headers map[string]string
Client *http.Client
}
type fleetStatusResponse struct {
@@ -41,7 +42,11 @@ func fleetReadinessHandler(client *http.Client, preferred capability.FleetTier,
for name, value := range endpoint.Headers {
request.Header.Set(name, value)
}
response, requestErr := client.Do(request)
probeClient := endpoint.Client
if probeClient == nil {
probeClient = client
}
response, requestErr := probeClient.Do(request)
if requestErr == nil {
response.Body.Close()
if response.StatusCode >= 200 && response.StatusCode < 300 {
+12 -7
View File
@@ -85,21 +85,26 @@ 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"
mux.Handle("/readyz", health.ReadinessHandler{Runtime: "cognitive-worker", Probes: []health.Probe{
health.HTTPProbe(readinessClient, strings.TrimRight(envOr("CORE_RUNTIME_URL", "http://core-runtime:8080"), "/")+"/healthz"),
health.OptionalHTTPProbe(readinessClient, strings.TrimRight(envOr("LOCAL_CPU_PROVIDER_URL", "http://docker.cin.su:18082/v1"), "/")+"/models"),
health.HTTPProbe(readinessClient, lifeboatURL),
health.OptionalHTTPProbe(readinessClient, strings.TrimRight(envOr("LOCAL_GPU_PROVIDER_URL", "http://192.168.220.91:11434"), "/")+"/api/tags"),
}})
fleetClient := &http.Client{Timeout: 2 * time.Second}
preferredFleet := capability.FleetTier(envOr("COGNITIVE_PREFERRED_FLEET_TIER", string(capability.FleetTierHeavy)))
cpuFleetHeaders := map[string]string{}
if cpuAPIKey != "" {
cpuFleetHeaders["Authorization"] = "Bearer " + cpuAPIKey
}
cpuEndpoint := strings.TrimRight(envOr("LOCAL_CPU_PROVIDER_URL", "http://docker.cin.su:18082/v1"), "/") + "/models"
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"),
cpuHealthProbe,
health.HTTPProbe(readinessClient, lifeboatURL),
health.OptionalHTTPProbe(readinessClient, strings.TrimRight(envOr("LOCAL_GPU_PROVIDER_URL", "http://192.168.220.91:11434"), "/")+"/api/tags"),
}
mux.Handle("/readyz", health.ReadinessHandler{Runtime: "cognitive-worker", Probes: readinessProbes})
mux.Handle("/v1/fleet/readiness", fleetReadinessHandler(fleetClient, preferredFleet, []fleetEndpoint{
{ProviderID: "local-cpu-primary", ModelID: "local-cpu-model", Tier: capability.FleetTierFast, URL: strings.TrimRight(envOr("LOCAL_CPU_PROVIDER_URL", "http://docker.cin.su:18082/v1"), "/") + "/models", Headers: cpuFleetHeaders},
{ProviderID: "local-cpu-primary", ModelID: "local-cpu-model", Tier: capability.FleetTierFast, URL: cpuEndpoint, Headers: cpuFleetHeaders, Client: &http.Client{Timeout: 10 * time.Second}},
{ProviderID: "local-medium-unconfigured", ModelID: "medium-unconfigured", Tier: capability.FleetTierMedium},
{ProviderID: "local-gpu-opportunistic", ModelID: "gpu-opportunistic", Tier: capability.FleetTierHeavy, URL: strings.TrimRight(envOr("LOCAL_GPU_PROVIDER_URL", "http://192.168.220.91:11434"), "/") + "/api/tags"},
{ProviderID: "cognitive-lifeboat", ModelID: "qwen3-8b-lifeboat", Tier: capability.FleetTierLifeboat, URL: lifeboatURL},
+2 -2
View File
@@ -1,5 +1,5 @@
{
"task_file": "dev/tasks/DEV-000164-authorized-fleet-readiness.json",
"task_id": "DEV-000164",
"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-000165",
"task_revision": 1,
"architecture_revision": "ADR-0018",
"base_repo_revision": "5d5afc88883c7f79e3b051c790642e41386666ec",
"risk_tier": "E",
"intent": "Use the existing file-backed CPU provider API key in cognitive-worker readiness probes so /readyz and fleet readiness report consistent provider state.",
"requirement_ids": ["WP-601", "Model fleet/failover"],
"allowed_change_scope": {"modules": ["cognitive-worker", "platform-health"], "paths": ["runtime/health/server.go", "runtime/health/server_test.go", "cmd/cognitive-worker/main.go", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000165-authorized-cpu-health-probe.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 health Probe contract", "existing provider API key resolver"], "may_change": ["health probe header support"]},
"semantic_owner": "cognitive-worker/platform health",
"canonical_state": [],
"state_transitions": ["authenticated CPU probe -> consistent readiness evidence"],
"side_effects_and_privileges": ["resolved key is used only in outbound probe header; never logged or persisted"],
"persistence_and_migrations": ["none"],
"failure_retry_restart": ["provider timeout or rejected key remains degraded; no fallback behavior changed"],
"assumptions": ["the local provider uses Bearer authentication"],
"acceptance_scenarios": ["health probe sends configured header", "cognitive /readyz uses authenticated CPU probe", "GPU remains optional", "no key is sent when absent"],
"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"]
}
@@ -0,0 +1,23 @@
{
"task_id": "DEV-000166",
"task_revision": 1,
"architecture_revision": "ADR-0018",
"base_repo_revision": "4086b7522460a78afc341d996ca7874505018871",
"risk_tier": "E",
"intent": "Give the authenticated CPU readiness probe the provider adapter timeout while retaining short opportunistic GPU probing.",
"requirement_ids": ["WP-601", "Model fleet/failover"],
"allowed_change_scope": {"modules": ["cognitive-worker", "platform-health"], "paths": ["runtime/health/server.go", "runtime/health/server_test.go", "cmd/cognitive-worker/main.go", "cmd/cognitive-worker/fleet_status.go", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000166-per-provider-readiness-timeouts.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 health Probe contract", "existing fleet readiness probe"], "may_change": ["per-probe timeout metadata"]},
"semantic_owner": "cognitive-worker/platform health",
"canonical_state": [],
"state_transitions": ["provider-specific probe deadline -> truthful readiness result"],
"side_effects_and_privileges": ["no new privilege; existing provider auth only"],
"persistence_and_migrations": ["none"],
"failure_retry_restart": ["CPU gets bounded 10-second probe; GPU retains 2-second opportunistic timeout"],
"assumptions": ["CPU provider catalog may take longer than opportunistic probes while remaining within adapter bounds"],
"acceptance_scenarios": ["per-probe health timeout is honored", "CPU fleet client uses 10 seconds", "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"]
}
@@ -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"]
}
+19 -1
View File
@@ -29,6 +29,7 @@ type Probe struct {
Name string
Check func(context.Context) error
Optional bool
Timeout time.Duration
}
type Result struct {
@@ -56,7 +57,11 @@ func (h ReadinessHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
if probe.Check == nil {
result.State, result.Reason, result.Detail = "DEGRADED", "PROBE_NOT_CONFIGURED", "probe is not configured"
} else {
ctx, cancel := context.WithTimeout(r.Context(), timeout)
probeTimeout := timeout
if probe.Timeout > 0 {
probeTimeout = probe.Timeout
}
ctx, cancel := context.WithTimeout(r.Context(), probeTimeout)
err := probe.Check(ctx)
cancel()
if err != nil {
@@ -82,11 +87,18 @@ func (h ReadinessHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
}
func HTTPProbe(client *http.Client, endpoint string) Probe {
return HTTPProbeWithHeaders(client, endpoint, nil)
}
func HTTPProbeWithHeaders(client *http.Client, endpoint string, headers map[string]string) Probe {
return Probe{Name: endpoint, Check: func(ctx context.Context) error {
req, err := http.NewRequestWithContext(ctx, http.MethodGet, endpoint, nil)
if err != nil {
return err
}
for name, value := range headers {
req.Header.Set(name, value)
}
response, err := client.Do(req)
if err != nil {
return err
@@ -108,6 +120,12 @@ func OptionalHTTPProbe(client *http.Client, endpoint string) Probe {
return probe
}
func OptionalHTTPProbeWithHeaders(client *http.Client, endpoint string, headers map[string]string) Probe {
probe := HTTPProbeWithHeaders(client, endpoint, headers)
probe.Optional = true
return probe
}
func TCPProbe(name, host, port string) Probe {
return Probe{Name: name, Check: func(ctx context.Context) error {
connection, err := (&net.Dialer{}).DialContext(ctx, "tcp", net.JoinHostPort(host, port))
+26
View File
@@ -7,6 +7,7 @@ import (
"net/http/httptest"
"strings"
"testing"
"time"
)
func TestReadinessReturns503WithTypedDependencyReason(t *testing.T) {
@@ -27,6 +28,31 @@ func TestReadinessReturns200WhenAllProbesPass(t *testing.T) {
}
}
func TestHTTPProbeWithHeadersSendsHeaders(t *testing.T) {
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.Header.Get("Authorization") != "Bearer health-key" {
t.Fatalf("authorization=%q", r.Header.Get("Authorization"))
}
w.WriteHeader(http.StatusOK)
}))
defer server.Close()
probe := HTTPProbeWithHeaders(server.Client(), server.URL, map[string]string{"Authorization": "Bearer health-key"})
if err := probe.Check(context.Background()); err != nil {
t.Fatal(err)
}
}
func TestReadinessUsesPerProbeTimeout(t *testing.T) {
handler := ReadinessHandler{Runtime: "test", Timeout: time.Millisecond, Probes: []Probe{{
Name: "slow-but-required", Timeout: time.Second, Check: func(context.Context) error { return nil },
}}}
recorder := httptest.NewRecorder()
handler.ServeHTTP(recorder, httptest.NewRequest(http.MethodGet, "/readyz", nil))
if recorder.Code != http.StatusOK {
t.Fatalf("readiness response = %d %s", recorder.Code, recorder.Body.String())
}
}
func TestReadinessStaysAvailableWithOptionalDependencyDegraded(t *testing.T) {
handler := ReadinessHandler{Runtime: "test", Probes: []Probe{
{Name: "lifeboat", Check: func(context.Context) error { return nil }},