Preserve TUS sessions and latest retry errors

Keep successful upload creation sessions usable when the optional response body is interrupted, while still requiring Location. Preserve the latest failure across retries and add regression coverage and documentation.

README.md

@@ -467,6 +467,8 @@ retransmitted when a request fails.
 An HTTP 500 during resumable creation is a server/provider failure, not proof
 that the recording filename is invalid. Creation errors include up to 4 KiB of
 the server response, and the final retry error retains the latest failure cause.
+Created upload sessions remain usable if only the optional creation response
+body is interrupted; the successful response's `Location` is retained.
 Check the corresponding Hub/Vault logs and storage-provider configuration.
 As a temporary diagnostic workaround, set `AGENT_DISABLE_RESUMABLE_UPLOAD=true`
 and restart Agent to use ordinary uploads. This does not change the recording

machinery/src/cloud/tus_client.go

@@ -167,7 +167,6 @@ func runTusUpload(baseURL, metadata, fileName, label, slot string, setHeaders tu
 	var lastErr error
 
 	for attempt := 0; attempt < maxAttempts; attempt++ {
-		lastErr = nil
 		// (1) Ensure we have an active upload URL, creating one if needed.
 		if uploadURL == "" {
 			created, status, cerr := tusCreate(client, baseURL, size, metadata, setHeaders, fileName)
@@ -424,8 +423,9 @@ func tusCreate(client *http.Client, baseURL string, size int64, metadata string,
 		}
 		return "", resp.StatusCode, fmt.Errorf("unexpected status creating upload: %s", resp.Status)
 	}
+	// Location identifies the created upload; its response body is optional.
 	if _, err := io.Copy(io.Discard, resp.Body); err != nil {
-		return "", resp.StatusCode, fmt.Errorf("read upload creation response: %w", err)
+		log.WithError(err).WithField("component", "tus").Debug("Could not drain optional upload creation response body")
 	}
 	location := resp.Header.Get("Location")
 	if location == "" {

machinery/src/cloud/tus_client_test.go

@@ -22,6 +22,80 @@ import (
 	"github.com/kerberos-io/agent/machinery/src/utils"
 )
 
+func TestResumableEOFRestartPreservesLatestFailure(t *testing.T) {
+	previousDelay := tusBackoffBaseDelay
+	tusBackoffBaseDelay = 0
+	t.Cleanup(func() { tusBackoffBaseDelay = previousDelay })
+	const fileName = "1767355200123-device-key.mp4"
+	withRecording(t, fileName, []byte("recording"))
+	fake := newFakeTus()
+	fake.failFinalize = 1
+	var heads atomic.Int32
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.Method == http.MethodHead && heads.Add(1) <= 2 {
+			w.WriteHeader(http.StatusServiceUnavailable)
+			return
+		}
+		fake.ServeHTTP(w, r)
+	}))
+	defer server.Close()
+
+	uploaded, responded, supported, _, err := uploadVaultResumable(testVault(server.URL), "public", "device-key", fileName, "test", "primary")
+	if uploaded || !responded || !supported || err == nil ||
+		!strings.Contains(err.Error(), "unexpected status on PATCH: 502 Bad Gateway") ||
+		strings.Contains(err.Error(), "503 Service Unavailable") {
+		t.Fatalf("upload = %t/%t/%t, %v; want the latest finalization failure after EOF restart", uploaded, responded, supported, err)
+	}
+	patches, _ := fake.patchCounts()
+	if heads.Load() != 4 || fake.createCount() != 1 || patches != 1 ||
+		len(fake.requestsForMethod(http.MethodDelete)) != 1 {
+		t.Fatalf("unexpected retry sequence: heads=%d, creates=%d, patches=%d", heads.Load(), fake.createCount(), patches)
+	}
+}
+
+func TestResumableCreationKeepsLocationAfterBodyReadError(t *testing.T) {
+	previousDelay := tusBackoffBaseDelay
+	tusBackoffBaseDelay = 0
+	t.Cleanup(func() { tusBackoffBaseDelay = previousDelay })
+	const fileName = "1767355200123-device-key.mp4"
+	payload := []byte("recording")
+	withRecording(t, fileName, payload)
+	fake := newFakeTus()
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.Method == http.MethodPost {
+			w.Header().Set("Content-Length", "100")
+			fake.ServeHTTP(w, r)
+			_, _ = io.WriteString(w, "short")
+			return
+		}
+		fake.ServeHTTP(w, r)
+	}))
+	defer server.Close()
+
+	uploaded, responded, supported, _, err := uploadVaultResumable(testVault(server.URL), "public", "device-key", fileName, "test", "primary")
+	if !uploaded || !responded || !supported || err != nil {
+		t.Fatalf("upload = %t/%t/%t, %v; want success despite the interrupted optional creation body", uploaded, responded, supported, err)
+	}
+	if fake.createCount() != 1 || fake.totalBytes() != int64(len(payload)) {
+		t.Fatalf("created %d slots and uploaded %d bytes; want one reused slot and %d bytes", fake.createCount(), fake.totalBytes(), len(payload))
+	}
+}
+
+func TestTusCreateStillRequiresLocationAfterBodyReadError(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Length", "100")
+		w.WriteHeader(http.StatusCreated)
+		_, _ = io.WriteString(w, "short")
+	}))
+	defer server.Close()
+	location, status, err := tusCreate(server.Client(), server.URL, 8, "",
+		func(http.Header, string) {}, "1767355200123-device-key.mp4")
+	if location != "" || status != http.StatusCreated || err == nil ||
+		!strings.Contains(err.Error(), "missing Location") {
+		t.Fatalf("creation = %q, %d, %v; want missing Location error", location, status, err)
+	}
+}
+
 func TestTusCreateReportsBoundedServerErrors(t *testing.T) {
 	for _, tc := range []struct {
 		name string