hub v0.38.1: offsite provisioning must survive a client disconnect (live F1)

First supervised live run: the ~25s spinner-less offsite save invited a
re-click; the abandoned first request's r.Context() was canceled between
CreateSubaccount and SaveOneTimeSecret, stranding sub-account 268985 with a
password lost forever (consume 404s permanently).

applyOffsite now provisions on context.WithoutCancel + 3-minute absolute
timeout: once the create starts, create->wait->store runs to completion.
Regression test with a ctx-honoring fake that cancels the request context
mid-create; red-proofed against the raw-ctx pre-fix shape (reproduces the
exact live error).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PSK5g6qYLknKj8u3QAFEr6

hub/internal/web/configs.go

@@ -817,6 +817,13 @@ func (s *Server) applyOffsite(ctx context.Context, r *http.Request, cfg *store.C
 	if s.offsite == nil {
 		return fmt.Errorf("offsite provisioning is not configured on this hub (no Hetzner token)")
 	}
+	// Detach from the client's request context: provisioning takes ~25s (create + wait + host-key scan) and
+	// an impatient re-click cancels r.Context() MID-SEQUENCE — live finding: the cancel landed between
+	// CreateSubaccount and SaveOneTimeSecret, stranding a sub-account whose one-time password was lost
+	// forever. Once provisioning starts it must run to completion (create→wait→store is the atom); the
+	// absolute timeout still bounds a hung Hetzner call.
+	ctx, cancel := context.WithTimeout(context.WithoutCancel(ctx), 3*time.Minute)
+	defer cancel()
 	in := offsite.Input{
 		Enabled: true,
 		Type:    strings.TrimSpace(r.FormValue("offsite_type")),

hub/internal/web/configs_offsite_test.go

@@ -0,0 +1,89 @@
+package web
+
+import (
+	"context"
+	"io"
+	"log"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-hub/internal/hetznerapi"
+	"gitea.dooplex.hu/admin/felhom-hub/internal/offsite"
+	"gitea.dooplex.hu/admin/felhom-hub/internal/store"
+)
+
+// disconnectingAPI models the live F1 failure (VALIDATION-offsite-provisioning-e2e 2026-07-09): the
+// browser disconnects (impatient re-click) while the subaccount create action is in flight — the request
+// context is canceled MID-SEQUENCE. Every call honors the ctx it receives, like the real HTTP client.
+type disconnectingAPI struct {
+	*hetznerapi.Fake
+	cancelReq context.CancelFunc
+}
+
+func (a *disconnectingAPI) CreateSubaccount(ctx context.Context, boxID int64, req hetznerapi.CreateSubaccountRequest) (int64, hetznerapi.Action, error) {
+	if err := ctx.Err(); err != nil {
+		return 0, hetznerapi.Action{}, err
+	}
+	id, act, err := a.Fake.CreateSubaccount(ctx, boxID, req)
+	a.cancelReq() // the client disconnects while the create action runs on Hetzner's side
+	return id, act, err
+}
+
+func (a *disconnectingAPI) WaitAction(ctx context.Context, act hetznerapi.Action) error {
+	if err := ctx.Err(); err != nil {
+		return err
+	}
+	return a.Fake.WaitAction(ctx, act)
+}
+
+func (a *disconnectingAPI) GetSubaccount(ctx context.Context, boxID, subID int64) (hetznerapi.Subaccount, error) {
+	if err := ctx.Err(); err != nil {
+		return hetznerapi.Subaccount{}, err
+	}
+	return a.Fake.GetSubaccount(ctx, boxID, subID)
+}
+
+type webTestScanner struct{}
+
+func (webTestScanner) Fingerprint(_ context.Context, _ string, _ int) (string, error) {
+	return "SHA256:webtestfp", nil
+}
+
+// Live finding F1: a re-click canceled r.Context() between CreateSubaccount and SaveOneTimeSecret —
+// the sub-account was created on Hetzner but its one-time password was lost forever (stranded resource,
+// controller consume 404s permanently). applyOffsite must detach provisioning from the client's context:
+// once the create starts, create→wait→store runs to completion even if the client goes away.
+func TestApplyOffsite_ClientDisconnectMidProvision(t *testing.T) {
+	s, st := newTestServer(t)
+	reqCtx, cancel := context.WithCancel(context.Background())
+	api := &disconnectingAPI{Fake: hetznerapi.NewFake(), cancelReq: cancel}
+	s.SetOffsiteProvisioner(&offsite.Provisioner{
+		API: api, Store: st, Scanner: webTestScanner{}, PoolBoxID: 611714, Location: "fsn1",
+		Logger: log.New(io.Discard, "", 0),
+	})
+
+	r := httptest.NewRequest(http.MethodPost,
+		"/configs/cust-web/edit?offsite_enabled=on&offsite_type=shared&offsite_quota_gb=50", nil)
+	r = r.WithContext(reqCtx)
+	cfg := &store.CustomerConfig{CustomerID: "cust-web", ConfigJSON: "{}"}
+
+	err := s.applyOffsite(r.Context(), r, cfg)
+
+	if reqCtx.Err() == nil {
+		t.Fatal("test harness broken: the request context was never canceled (no disconnect simulated)")
+	}
+	if err != nil {
+		t.Fatalf("provisioning must survive a client disconnect mid-sequence, got: %v", err)
+	}
+	// The one-time password reached the store — the whole point: no stranded resource with a lost password.
+	pw, err := st.ConsumeOneTimeSecret("cust-web")
+	if err != nil || pw == "" {
+		t.Fatalf("one-time password must be stored despite the disconnect (else the sub-account is stranded): pw=%q err=%v", pw, err)
+	}
+	// And the descriptor (incl. the pin) was merged for the caller to save.
+	if !strings.Contains(cfg.ConfigJSON, `"offsite"`) || !strings.Contains(cfg.ConfigJSON, "SHA256:webtestfp") {
+		t.Fatalf("descriptor not merged into ConfigJSON: %s", cfg.ConfigJSON)
+	}
+}