feat(coordinator): complete tasks with an artifact_id, not a URI (CTX-05, part 3)

Task results are now coordinator-owned artifacts end to end. - domain.Task carries ResultArtifactID instead of ResultURI/ResultSHA256; CompleteWith and its idempotency key are keyed on the artifact id. - CompleteTask verifies the referenced artifact was stored for this exact task (rule 10): a worker cannot finish task B with task A's artifact, nor name an id that isn't a partial_result. Mismatch → 409. - POST /tasks/{id}/result takes {result:{artifact_id,...}}; ListResults and ResultManifest follow. - migration 0004 drops result_uri/result_sha256 and requires a completed task to reference its result_artifact_id. - smoke and requests.http exercise upload → complete-by-id → replay → conflict.
2026-07-23 15:54:11 +03:00
parent 6d45406ee0
commit 58da6ef139
14 changed files with 184 additions and 116 deletions
@@ -28,27 +28,26 @@ const ErrCodeLeaseExpired = "lease_expired"
 // Nullable columns are pointers so "no lease" stays distinguishable from
 // "lease owned by the empty string" — a plain string cannot express both.
 type Task struct {
-	ID             uuid.UUID
-	JobID          uuid.UUID
-	ChunkIndex     int
-	Workload       string
-	InputURI       string
-	InputSHA256    string
-	Parameters     map[string]any
-	Status         TaskStatus
-	Attempt        int
-	MaxAttempts    int
-	LeaseOwner     *string
-	LeaseExpiresAt *time.Time
-	ResultURI      *string
-	ResultSHA256   *string
-	Metrics        map[string]any
-	ErrorCode      *string
-	ErrorMessage   *string
-	CreatedAt      time.Time
-	StartedAt      *time.Time
-	CompletedAt    *time.Time
-	Version        int
+	ID               uuid.UUID
+	JobID            uuid.UUID
+	ChunkIndex       int
+	Workload         string
+	InputURI         string
+	InputSHA256      string
+	Parameters       map[string]any
+	Status           TaskStatus
+	Attempt          int
+	MaxAttempts      int
+	LeaseOwner       *string
+	LeaseExpiresAt   *time.Time
+	ResultArtifactID *uuid.UUID
+	Metrics          map[string]any
+	ErrorCode        *string
+	ErrorMessage     *string
+	CreatedAt        time.Time
+	StartedAt        *time.Time
+	CompletedAt      *time.Time
+	Version          int
 }

 // NewTask builds a pending task. maxAttempts <= 0 falls back to the default.
@@ -146,17 +145,16 @@ func (t *Task) RenewLease(worker string, attempt int, until time.Time) error {
 // retry the same manifest, and that must succeed rather than trip the lease
 // check on a task the coordinator already finished. A *different* manifest for
 // an already-completed task is a genuine conflict.
-func (t *Task) CompleteWith(resultURI, resultSHA256 string, metrics map[string]any,
+func (t *Task) CompleteWith(resultArtifactID uuid.UUID, metrics map[string]any,
 	worker string, attempt int, now time.Time) error {

-	if resultURI == "" || resultSHA256 == "" {
+	if resultArtifactID == uuid.Nil {
 		return ErrInvalidInput
 	}

 	if t.Status == TaskCompleted {
-		if t.Attempt == attempt && t.ResultURI != nil && *t.ResultURI == resultURI &&
-			t.ResultSHA256 != nil && *t.ResultSHA256 == resultSHA256 {
-			return nil // same attempt, same manifest — replay of a successful call
+		if t.Attempt == attempt && t.ResultArtifactID != nil && *t.ResultArtifactID == resultArtifactID {
+			return nil // same attempt, same artifact — replay of a successful call
 		}
 		return ErrResultConflict
 	}
@@ -166,8 +164,7 @@ func (t *Task) CompleteWith(resultURI, resultSHA256 string, metrics map[string]a
 	}

 	t.Status = TaskCompleted
-	t.ResultURI = &resultURI
-	t.ResultSHA256 = &resultSHA256
+	t.ResultArtifactID = &resultArtifactID
 	t.Metrics = metrics
 	t.CompletedAt = &now
 	t.LeaseOwner = nil
@@ -234,11 +231,11 @@ type ClaimedTask struct {
 	LeaseExpiresAt time.Time
 }

-// ResultManifest is a completed task's output, ordered for the stitcher.
+// ResultManifest is a completed task's output, ordered for the stitcher. It
+// points at the coordinator-owned result artifact rather than a worker URI.
 type ResultManifest struct {
-	TaskID       uuid.UUID
-	ChunkIndex   int
-	ResultURI    string
-	ResultSHA256 string
-	Metrics      map[string]any
+	TaskID           uuid.UUID
+	ChunkIndex       int
+	ResultArtifactID uuid.UUID
+	Metrics          map[string]any
 }
@@ -9,9 +9,11 @@ import (
 )

 var (
-	testNow    = time.Date(2026, 7, 21, 12, 0, 0, 0, time.UTC)
-	testLater  = testNow.Add(time.Hour)
-	testWorker = "worker-1"
+	testNow       = time.Date(2026, 7, 21, 12, 0, 0, 0, time.UTC)
+	testLater     = testNow.Add(time.Hour)
+	testWorker    = "worker-1"
+	testResult    = uuid.New()
+	testResultAlt = uuid.New()
 )

 // leasedTask builds a task already leased to testWorker at the given attempt.
@@ -32,7 +34,7 @@ func leasedTask(attempt, maxAttempts int) *Task {
 func TestCompleteWithRecordsResult(t *testing.T) {
 	task := leasedTask(1, 3)

-	if err := task.CompleteWith("s3://r.csv", "abc", nil, testWorker, 1, testNow); err != nil {
+	if err := task.CompleteWith(testResult, nil, testWorker, 1, testNow); err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
 	if task.Status != TaskCompleted {
@@ -50,12 +52,12 @@ func TestCompleteWithRecordsResult(t *testing.T) {
 // rather than fail on the lease it has already given up.
 func TestCompleteWithIsIdempotentForSameManifest(t *testing.T) {
 	task := leasedTask(1, 3)
-	if err := task.CompleteWith("s3://r.csv", "abc", nil, testWorker, 1, testNow); err != nil {
+	if err := task.CompleteWith(testResult, nil, testWorker, 1, testNow); err != nil {
 		t.Fatalf("first call: %v", err)
 	}
 	versionAfterFirst := task.Version

-	if err := task.CompleteWith("s3://r.csv", "abc", nil, testWorker, 1, testLater); err != nil {
+	if err := task.CompleteWith(testResult, nil, testWorker, 1, testLater); err != nil {
 		t.Fatalf("replay must be idempotent, got %v", err)
 	}
 	if task.Version != versionAfterFirst {
@@ -65,11 +67,11 @@ func TestCompleteWithIsIdempotentForSameManifest(t *testing.T) {

 func TestCompleteWithRejectsDifferentManifest(t *testing.T) {
 	task := leasedTask(1, 3)
-	if err := task.CompleteWith("s3://r.csv", "abc", nil, testWorker, 1, testNow); err != nil {
+	if err := task.CompleteWith(testResult, nil, testWorker, 1, testNow); err != nil {
 		t.Fatalf("first call: %v", err)
 	}

-	err := task.CompleteWith("s3://other.csv", "def", nil, testWorker, 1, testLater)
+	err := task.CompleteWith(testResultAlt, nil, testWorker, 1, testLater)
 	if !errors.Is(err, ErrResultConflict) {
 		t.Errorf("err = %v, want ErrResultConflict", err)
 	}
@@ -78,7 +80,7 @@ func TestCompleteWithRejectsDifferentManifest(t *testing.T) {
 func TestCompleteWithRejectsForeignWorker(t *testing.T) {
 	task := leasedTask(1, 3)

-	err := task.CompleteWith("s3://r.csv", "abc", nil, "worker-2", 1, testNow)
+	err := task.CompleteWith(testResult, nil, "worker-2", 1, testNow)
 	if !errors.Is(err, ErrLeaseConflict) {
 		t.Errorf("err = %v, want ErrLeaseConflict", err)
 	}
@@ -87,7 +89,7 @@ func TestCompleteWithRejectsForeignWorker(t *testing.T) {
 func TestCompleteWithRejectsStaleAttempt(t *testing.T) {
 	task := leasedTask(2, 3) // task is on attempt 2

-	err := task.CompleteWith("s3://r.csv", "abc", nil, testWorker, 1, testNow) // worker thinks it is 1
+	err := task.CompleteWith(testResult, nil, testWorker, 1, testNow) // worker thinks it is 1
 	if !errors.Is(err, ErrStaleAttempt) {
 		t.Errorf("err = %v, want ErrStaleAttempt", err)
 	}