feat: deliver Sense audits to Bell (T-016)
Harness governance / validate (push) Has been cancelled
Harness governance / validate (pull_request) Has been cancelled

This commit is contained in:
QiuSW
2026-08-11 00:24:32 +08:00
parent aca22f4667
commit 4be2386421
35 changed files with 2188 additions and 42 deletions
@@ -0,0 +1,132 @@
package auditrelay
import (
"context"
"crypto/sha256"
"encoding/base64"
"encoding/json"
"errors"
"net/http"
"net/http/httptest"
"os"
"path/filepath"
"testing"
"time"
)
func testEnvelope(id string) Envelope {
return Envelope{SchemaVersion: 1, Event: Event{
EventID: id, EventType: "device.created", TenantID: "tenant", SiteID: "site", DeviceID: "camera-1",
Actor: Actor{Type: "system", ID: "sense"}, AggregateGeneration: 1,
ProjectionVersions: ProjectionVersions{}, Data: json.RawMessage(`{"kind":"device_created"}`), OccurredAt: time.Unix(1, 0).UTC(),
}}
}
func TestClientSignsCanonicalRequest(t *testing.T) {
secret := bytesOf(32, 7)
server := httptest.NewServer(http.HandlerFunc(func(writer http.ResponseWriter, request *http.Request) {
body := make([]byte, request.ContentLength)
_, _ = request.Body.Read(body)
canonical := CanonicalString(request.Method, request.URL.Path, request.Header.Get(HeaderTimestamp), request.Header.Get(HeaderNonce), body)
if request.Header.Get(HeaderSignature) != Signature(secret, canonical) {
t.Error("request signature did not match canonical vector")
}
_ = json.NewEncoder(writer).Encode(BatchResponse{Results: []Result{{EventID: "audit_00000000000000000000000000000001", Status: "accepted"}}})
}))
defer server.Close()
client, err := NewClient(server.URL+RelayPath, "sense-a", secret, server.Client())
if err != nil {
t.Fatal(err)
}
client.now = func() time.Time { return time.Unix(1_800_000_000, 0) }
client.nonce = func() (string, error) { return "AAAAAAAAAAAAAAAAAAAAAA", nil }
results, err := client.Send(context.Background(), []Envelope{testEnvelope("audit_00000000000000000000000000000001")})
if err != nil || len(results) != 1 || results[0].Status != "accepted" {
t.Fatalf("unexpected relay result: %+v, %v", results, err)
}
}
func TestEndpointAndKeySecurity(t *testing.T) {
if _, err := ValidateEndpoint("http://example.com" + RelayPath); err == nil {
t.Fatal("remote plaintext relay URL was accepted")
}
if _, err := ValidateEndpoint("https://example.com" + RelayPath + "?secret=x"); err == nil {
t.Fatal("relay URL query was accepted")
}
secret := bytesOf(32, 9)
path := filepath.Join(t.TempDir(), "keys.json")
document := map[string]any{"version": 1, "keys": []any{map[string]any{"key_id": "sense-a", "secret_base64url": base64.RawURLEncoding.EncodeToString(secret)}}}
raw, _ := json.Marshal(document)
if err := os.WriteFile(path, raw, 0o600); err != nil {
t.Fatal(err)
}
loaded, err := LoadKey(path, "sense-a")
if err != nil || sha256.Sum256(loaded) != sha256.Sum256(secret) {
t.Fatalf("external key was not loaded: %v", err)
}
}
type fakeRepository struct {
queued []QueuedEvent
completed []Completion
}
func (*fakeRepository) AuditRelayReady(context.Context) error { return nil }
func (f *fakeRepository) ClaimAuditRelayBatch(context.Context, string, int, time.Duration) ([]QueuedEvent, error) {
return f.queued, nil
}
func (f *fakeRepository) CompleteAuditRelayBatch(_ context.Context, _ string, values []Completion) error {
f.completed = append([]Completion(nil), values...)
return nil
}
type fakeSender struct {
results []Result
err error
}
func (f fakeSender) Send(context.Context, []Envelope) ([]Result, error) { return f.results, f.err }
func TestWorkerDispositionAndBackoff(t *testing.T) {
first := "audit_00000000000000000000000000000001"
second := "audit_00000000000000000000000000000002"
repository := &fakeRepository{queued: []QueuedEvent{{Envelope: testEnvelope(first), LeaseToken: 1, AttemptCount: 1}, {Envelope: testEnvelope(second), LeaseToken: 2, AttemptCount: 10}}}
code := "schema_invalid"
worker, _ := NewWorker(repository, fakeSender{results: []Result{{EventID: first, Status: "accepted"}, {EventID: second, Status: "rejected", ErrorCode: &code}}}, "worker")
if err := worker.RelayOnce(context.Background()); err != nil {
t.Fatal(err)
}
if repository.completed[0].Disposition != Delivered || repository.completed[1].Disposition != DeadLetter {
t.Fatalf("unexpected dispositions: %+v", repository.completed)
}
repository.completed = nil
worker.sender = fakeSender{err: errors.New("network")}
if err := worker.RelayOnce(context.Background()); err == nil {
t.Fatal("network failure was hidden")
}
if repository.completed[0].RetryAfter != time.Second || repository.completed[1].RetryAfter != 300*time.Second {
t.Fatalf("retry bounds drifted: %+v", repository.completed)
}
}
func TestWorkerRetriesWholeBatchForIncompleteResponse(t *testing.T) {
first := "audit_00000000000000000000000000000001"
second := "audit_00000000000000000000000000000002"
repository := &fakeRepository{queued: []QueuedEvent{{Envelope: testEnvelope(first), LeaseToken: 1, AttemptCount: 1}, {Envelope: testEnvelope(second), LeaseToken: 2, AttemptCount: 2}}}
worker, _ := NewWorker(repository, fakeSender{results: []Result{{EventID: first, Status: "accepted"}}}, "worker")
if err := worker.RelayOnce(context.Background()); err == nil {
t.Fatal("incomplete response was accepted")
}
if len(repository.completed) != 2 || repository.completed[0].Disposition != Retry || repository.completed[1].Disposition != Retry {
t.Fatalf("incomplete response partially completed the batch: %+v", repository.completed)
}
}
func bytesOf(size int, value byte) []byte {
result := make([]byte, size)
for index := range result {
result[index] = value
}
return result
}
+131
View File
@@ -0,0 +1,131 @@
package auditrelay
import (
"bytes"
"context"
"crypto/hmac"
"crypto/rand"
"crypto/sha256"
"encoding/base64"
"encoding/hex"
"encoding/json"
"errors"
"fmt"
"io"
"net"
"net/http"
"net/url"
"strconv"
"strings"
"time"
)
const (
HeaderKeyID = "X-YoVision-Key-Id"
HeaderTimestamp = "X-YoVision-Timestamp"
HeaderNonce = "X-YoVision-Nonce"
HeaderSignature = "X-YoVision-Signature"
RelayPath = "/internal/v1/audit-events:batch"
)
type Client struct {
endpoint *url.URL
keyID string
secret []byte
httpClient *http.Client
now func() time.Time
nonce func() (string, error)
}
func NewClient(rawURL, keyID string, secret []byte, client *http.Client) (*Client, error) {
endpoint, err := ValidateEndpoint(rawURL)
if err != nil {
return nil, err
}
if keyID == "" || len(secret) < 32 {
return nil, errors.New("audit relay key ID and 32-byte secret are required")
}
if client == nil {
client = &http.Client{Timeout: 10 * time.Second}
}
return &Client{endpoint: endpoint, keyID: keyID, secret: append([]byte(nil), secret...), httpClient: client, now: time.Now, nonce: randomNonce}, nil
}
func ValidateEndpoint(rawURL string) (*url.URL, error) {
parsed, err := url.Parse(rawURL)
if err != nil || parsed.Host == "" || parsed.Path != RelayPath || parsed.RawQuery != "" || parsed.Fragment != "" || parsed.User != nil {
return nil, errors.New("invalid Bell audit relay URL")
}
host := parsed.Hostname()
ip := net.ParseIP(host)
loopback := strings.EqualFold(host, "localhost") || (ip != nil && ip.IsLoopback())
if parsed.Scheme != "https" && !(parsed.Scheme == "http" && loopback) {
return nil, errors.New("Bell audit relay URL requires HTTPS outside loopback")
}
return parsed, nil
}
func CanonicalString(method, path, timestamp, nonce string, body []byte) string {
digest := sha256.Sum256(body)
return strings.Join([]string{method, path, timestamp, nonce, hex.EncodeToString(digest[:])}, "\n")
}
func Signature(secret []byte, canonical string) string {
mac := hmac.New(sha256.New, secret)
_, _ = mac.Write([]byte(canonical))
return base64.RawURLEncoding.EncodeToString(mac.Sum(nil))
}
func randomNonce() (string, error) {
value := make([]byte, 16)
if _, err := rand.Read(value); err != nil {
return "", err
}
return base64.RawURLEncoding.EncodeToString(value), nil
}
func (c *Client) Send(ctx context.Context, events []Envelope) ([]Result, error) {
if len(events) < 1 || len(events) > MaxBatchSize {
return nil, errors.New("audit relay batch must contain 1 to 100 events")
}
body, err := json.Marshal(BatchRequest{Events: events})
if err != nil || len(body) > MaxBodyBytes {
return nil, errors.New("encode audit relay batch")
}
requestContext, cancel := context.WithTimeout(ctx, 10*time.Second)
defer cancel()
timestamp := strconv.FormatInt(c.now().UTC().Unix(), 10)
nonce, err := c.nonce()
if err != nil {
return nil, errors.New("generate audit relay nonce")
}
request, err := http.NewRequestWithContext(requestContext, http.MethodPost, c.endpoint.String(), bytes.NewReader(body))
if err != nil {
return nil, errors.New("create audit relay request")
}
request.Header.Set("Content-Type", "application/json")
request.Header.Set(HeaderKeyID, c.keyID)
request.Header.Set(HeaderTimestamp, timestamp)
request.Header.Set(HeaderNonce, nonce)
request.Header.Set(HeaderSignature, Signature(c.secret, CanonicalString(http.MethodPost, RelayPath, timestamp, nonce, body)))
response, err := c.httpClient.Do(request)
if err != nil {
return nil, errors.New("send audit relay request")
}
defer response.Body.Close()
if response.StatusCode != http.StatusOK {
_, _ = io.Copy(io.Discard, io.LimitReader(response.Body, 4096))
return nil, fmt.Errorf("Bell audit relay returned HTTP %d", response.StatusCode)
}
decoder := json.NewDecoder(io.LimitReader(response.Body, MaxBodyBytes+1))
decoder.DisallowUnknownFields()
var decoded BatchResponse
if err := decoder.Decode(&decoded); err != nil {
return nil, errors.New("decode audit relay response")
}
var trailing any
if err := decoder.Decode(&trailing); !errors.Is(err, io.EOF) {
return nil, errors.New("audit relay response contains trailing data")
}
return decoded.Results, nil
}
+53
View File
@@ -0,0 +1,53 @@
package auditrelay
import (
"bytes"
"encoding/base64"
"encoding/json"
"errors"
"io"
"os"
"regexp"
)
var keyIDPattern = regexp.MustCompile(`^[A-Za-z0-9][A-Za-z0-9._-]{0,63}$`)
type keyFile struct {
Version int `json:"version"`
Keys []struct {
KeyID string `json:"key_id"`
Secret string `json:"secret_base64url"`
} `json:"keys"`
}
func LoadKey(path, keyID string) ([]byte, error) {
raw, err := os.ReadFile(path)
if err != nil {
return nil, errors.New("read audit relay key file")
}
var document keyFile
decoder := json.NewDecoder(bytes.NewReader(raw))
decoder.DisallowUnknownFields()
if err := decoder.Decode(&document); err != nil || document.Version != 1 || len(document.Keys) == 0 {
return nil, errors.New("invalid audit relay key file")
}
var trailing any
if err := decoder.Decode(&trailing); !errors.Is(err, io.EOF) {
return nil, errors.New("invalid audit relay key file")
}
values := make(map[string][]byte, len(document.Keys))
for _, candidate := range document.Keys {
secret, err := base64.RawURLEncoding.DecodeString(candidate.Secret)
if err != nil || !keyIDPattern.MatchString(candidate.KeyID) || len(secret) < 32 {
return nil, errors.New("invalid audit relay secret")
}
if _, duplicate := values[candidate.KeyID]; duplicate {
return nil, errors.New("duplicate audit relay key ID")
}
values[candidate.KeyID] = secret
}
if secret, exists := values[keyID]; exists {
return secret, nil
}
return nil, errors.New("audit relay key ID not found")
}
+89
View File
@@ -0,0 +1,89 @@
// Package auditrelay delivers Sense-owned audit facts to Bell without sharing databases.
package auditrelay
import (
"context"
"encoding/json"
"errors"
"time"
)
const (
MaxBatchSize = 100
MaxBodyBytes = 1 << 20
LeaseDuration = 30 * time.Second
)
type Actor struct {
Type string `json:"type"`
ID string `json:"id"`
}
type ProjectionVersions struct {
QuotaSourceVersion *int64 `json:"quota_source_version"`
AreaPolicySourceVersion *int64 `json:"area_policy_source_version"`
}
type Event struct {
EventID string `json:"event_id"`
EventType string `json:"event_type"`
TenantID string `json:"tenant_id"`
SiteID string `json:"site_id"`
DeviceID string `json:"device_id"`
Actor Actor `json:"actor"`
Reason *string `json:"reason"`
TraceID *string `json:"trace_id"`
AggregateGeneration int64 `json:"aggregate_generation"`
ProjectionVersions ProjectionVersions `json:"projection_versions"`
Data json.RawMessage `json:"data"`
OccurredAt time.Time `json:"occurred_at"`
}
type Envelope struct {
SchemaVersion int `json:"schema_version"`
Event Event `json:"event"`
}
type BatchRequest struct {
Events []Envelope `json:"events"`
}
type Result struct {
EventID string `json:"event_id"`
Status string `json:"status"`
ErrorCode *string `json:"error_code,omitempty"`
}
type BatchResponse struct {
Results []Result `json:"results"`
}
type QueuedEvent struct {
Envelope
LeaseToken int64
AttemptCount int
}
type Disposition string
const (
Delivered Disposition = "delivered"
DeadLetter Disposition = "dead_letter"
Retry Disposition = "retry"
)
type Completion struct {
EventID string
LeaseToken int64
Disposition Disposition
ErrorCode string
RetryAfter time.Duration
}
var ErrLeaseLost = errors.New("audit relay lease lost")
type Repository interface {
AuditRelayReady(context.Context) error
ClaimAuditRelayBatch(context.Context, string, int, time.Duration) ([]QueuedEvent, error)
CompleteAuditRelayBatch(context.Context, string, []Completion) error
}
+126
View File
@@ -0,0 +1,126 @@
package auditrelay
import (
"context"
"errors"
"fmt"
"regexp"
"time"
)
var stableErrorCode = regexp.MustCompile(`^[a-z][a-z0-9_]{0,63}$`)
type Sender interface {
Send(context.Context, []Envelope) ([]Result, error)
}
type Worker struct {
repository Repository
sender Sender
owner string
}
func NewWorker(repository Repository, sender Sender, owner string) (*Worker, error) {
if repository == nil || sender == nil || owner == "" {
return nil, errors.New("audit relay worker dependencies are required")
}
return &Worker{repository: repository, sender: sender, owner: owner}, nil
}
func RetryDelay(attempt int) time.Duration {
if attempt < 1 {
attempt = 1
}
if attempt > 9 {
return 300 * time.Second
}
delay := time.Second << (attempt - 1)
if delay > 300*time.Second {
return 300 * time.Second
}
return delay
}
func (w *Worker) RelayOnce(ctx context.Context) error {
queued, err := w.repository.ClaimAuditRelayBatch(ctx, w.owner, MaxBatchSize, LeaseDuration)
if err != nil || len(queued) == 0 {
return err
}
events := make([]Envelope, len(queued))
for index := range queued {
events[index] = queued[index].Envelope
}
results, sendErr := w.sender.Send(ctx, events)
if sendErr != nil {
completions := make([]Completion, len(queued))
for index, value := range queued {
completions[index] = Completion{EventID: value.Event.EventID, LeaseToken: value.LeaseToken, Disposition: Retry, ErrorCode: "delivery_failed", RetryAfter: RetryDelay(value.AttemptCount)}
}
if err := w.repository.CompleteAuditRelayBatch(ctx, w.owner, completions); err != nil {
return err
}
return sendErr
}
if len(results) != len(queued) {
return w.retryAll(ctx, queued, "invalid_response")
}
expected := make(map[string]bool, len(queued))
for _, value := range queued {
expected[value.Event.EventID] = true
}
byID := make(map[string]Result, len(results))
for _, result := range results {
validStatus := ((result.Status == "accepted" || result.Status == "duplicate") && result.ErrorCode == nil) ||
(result.Status == "rejected" && result.ErrorCode != nil && stableErrorCode.MatchString(*result.ErrorCode))
if !expected[result.EventID] || !validStatus {
return w.retryAll(ctx, queued, "invalid_response")
}
if _, duplicate := byID[result.EventID]; duplicate {
return w.retryAll(ctx, queued, "invalid_response")
}
byID[result.EventID] = result
}
completions := make([]Completion, 0, len(queued))
for _, value := range queued {
result, ok := byID[value.Event.EventID]
if !ok {
return w.retryAll(ctx, queued, "invalid_response")
}
switch result.Status {
case "accepted", "duplicate":
completions = append(completions, Completion{EventID: value.Event.EventID, LeaseToken: value.LeaseToken, Disposition: Delivered})
case "rejected":
completions = append(completions, Completion{EventID: value.Event.EventID, LeaseToken: value.LeaseToken, Disposition: DeadLetter, ErrorCode: *result.ErrorCode})
}
}
return w.repository.CompleteAuditRelayBatch(ctx, w.owner, completions)
}
func (w *Worker) retryAll(ctx context.Context, queued []QueuedEvent, code string) error {
values := make([]Completion, len(queued))
for index, value := range queued {
values[index] = Completion{EventID: value.Event.EventID, LeaseToken: value.LeaseToken, Disposition: Retry, ErrorCode: code, RetryAfter: RetryDelay(value.AttemptCount)}
}
if err := w.repository.CompleteAuditRelayBatch(ctx, w.owner, values); err != nil {
return err
}
return fmt.Errorf("audit relay %s", code)
}
func (w *Worker) Run(ctx context.Context, interval time.Duration, report func(error)) {
if interval <= 0 {
interval = time.Second
}
for {
if err := w.RelayOnce(ctx); err != nil && ctx.Err() == nil && report != nil {
report(err)
}
timer := time.NewTimer(interval)
select {
case <-ctx.Done():
timer.Stop()
return
case <-timer.C:
}
}
}