fix(security): make scan report latency independent of scan history (MCP-2205)

internal/security/scanner/service.go

@@ -32,6 +32,7 @@ type Storage interface {
 	SaveScanJob(job *ScanJob) error
 	GetScanJob(id string) (*ScanJob, error)
 	ListScanJobs(serverName string) ([]*ScanJob, error)
+	ListScanJobMetas(serverName string) ([]*ScanJobMeta, error)
 	GetLatestScanJob(serverName string) (*ScanJob, error)
 	DeleteScanJob(id string) error
 	DeleteServerScanJobs(serverName string) error
@@ -1083,25 +1084,23 @@ func (s *Service) GetScanReportByJobID(ctx context.Context, jobID string) (*Aggr
 	agg.Pass1Complete = job.ScanPass == ScanPassSecurityScan && job.Status == ScanJobStatusCompleted
 	agg.Pass2Complete = job.ScanPass == ScanPassSupplyChainAudit && job.Status == ScanJobStatusCompleted
 
-	// If this is a Pass 1 job, try to find and merge companion Pass 2 results
+	// If this is a Pass 1 job, try to find and merge companion Pass 2 results.
+	// The companion is resolved via the lightweight scan-job index, so this does
+	// NOT deserialize the full per-server scan history (MCP-2205).
 	if job.ScanPass == ScanPassSecurityScan || job.ScanPass == 0 {
-		allJobs, _ := s.storage.ListScanJobs(job.ServerName)
-		for _, j := range allJobs {
-			if j.ScanPass == ScanPassSupplyChainAudit && j.Status == ScanJobStatusCompleted && j.StartedAt.After(job.StartedAt) {
-				pass2Reports, err := s.storage.ListScanReportsByJob(j.ID)
-				if err == nil {
-					for _, r := range pass2Reports {
-						for i := range r.Findings {
-							r.Findings[i].ScanPass = ScanPassSupplyChainAudit
-						}
+		if companionID := s.findCompanionPass2JobID(job); companionID != "" {
+			pass2Reports, err := s.storage.ListScanReportsByJob(companionID)
+			if err == nil {
+				for _, r := range pass2Reports {
+					for i := range r.Findings {
+						r.Findings[i].ScanPass = ScanPassSupplyChainAudit
 					}
-					allMerged := append(reports, pass2Reports...)
-					allMerged = deduplicatePass2Findings(allMerged)
-					agg = AggregateReportsWithJobStatus(job.ID, job.ServerName, allMerged, job)
-					agg.Pass1Complete = true
-					agg.Pass2Complete = true
 				}
-				break
+				allMerged := append(reports, pass2Reports...)
+				allMerged = deduplicatePass2Findings(allMerged)
+				agg = AggregateReportsWithJobStatus(job.ID, job.ServerName, allMerged, job)
+				agg.Pass1Complete = true
+				agg.Pass2Complete = true
 			}
 		}
 
@@ -1118,6 +1117,40 @@ func (s *Service) GetScanReportByJobID(ctx context.Context, jobID string) (*Aggr
 	return agg, nil
 }
 
+// findCompanionPass2JobID returns the ID of the Pass-2 (supply-chain audit) job
+// that companions the given Pass-1 job: the earliest completed Pass-2 job that
+// started after it. It reads the lightweight scan-job metadata index rather than
+// the full job records, so its cost is independent of scan-output size and the
+// report path no longer slows down as a server accrues scan history (MCP-2205).
+// Returns "" when no companion exists.
+func (s *Service) findCompanionPass2JobID(pass1 *ScanJob) string {
+	metas, err := s.storage.ListScanJobMetas(pass1.ServerName)
+	if err != nil {
+		s.logger.Warn("failed to list scan job metadata for companion lookup",
+			zap.String("server", pass1.ServerName),
+			zap.Error(err),
+		)
+		return ""
+	}
+
+	var best *ScanJobMeta
+	for _, m := range metas {
+		if m.ScanPass != ScanPassSupplyChainAudit || m.Status != ScanJobStatusCompleted {
+			continue
+		}
+		if !m.StartedAt.After(pass1.StartedAt) {
+			continue
+		}
+		if best == nil || m.StartedAt.Before(best.StartedAt) {
+			best = m
+		}
+	}
+	if best == nil {
+		return ""
+	}
+	return best.ID
+}
+
 // deduplicatePass2Findings removes Pass 2 findings that duplicate Pass 1 findings.
 // The dedup key is scanner + rule_id + title (not location, since paths may differ between passes).
 func deduplicatePass2Findings(reports []*ScanReport) []*ScanReport {
@@ -1158,38 +1191,51 @@ func deduplicatePass2Findings(reports []*ScanReport) []*ScanReport {
 // findLatestPassJobs finds the latest Pass 1 and Pass 2 jobs for a server.
 // Returns (pass1Job, pass2Job, error). At least one must be non-nil on success.
 func (s *Service) findLatestPassJobs(serverName string) (*ScanJob, *ScanJob, error) {
-	jobs, err := s.storage.ListScanJobs(serverName)
+	// Read lightweight metadata rather than full job records so this scales with
+	// neither scan-output size nor history depth: we deserialize at most the two
+	// jobs we actually return (MCP-2205).
+	metas, err := s.storage.ListScanJobMetas(serverName)
 	if err != nil {
 		// Surface the underlying I/O error so the caller can distinguish
 		// transient failures from "no records found".
-		return nil, nil, fmt.Errorf("list scan jobs for %s: %w", serverName, err)
+		return nil, nil, fmt.Errorf("list scan job metadata for %s: %w", serverName, err)
 	}
-	if len(jobs) == 0 {
+	if len(metas) == 0 {
 		return nil, nil, fmt.Errorf("%w: %s", errNoScans, serverName)
 	}
 
-	// Sort by start time descending (newest first)
-	sort.Slice(jobs, func(i, j int) bool {
-		return jobs[i].StartedAt.After(jobs[j].StartedAt)
-	})
-
-	var pass1Job, pass2Job *ScanJob
-	for _, j := range jobs {
-		if j.ScanPass == ScanPassSupplyChainAudit && pass2Job == nil {
-			pass2Job = j
-		} else if (j.ScanPass == ScanPassSecurityScan || j.ScanPass == 0) && pass1Job == nil {
+	// Pick the newest Pass-1 and Pass-2 job IDs by start time.
+	var pass1Meta, pass2Meta *ScanJobMeta
+	for _, m := range metas {
+		switch m.ScanPass {
+		case ScanPassSupplyChainAudit:
+			if pass2Meta == nil || m.StartedAt.After(pass2Meta.StartedAt) {
+				pass2Meta = m
+			}
+		case ScanPassSecurityScan, 0:
 			// ScanPass == 0 handles legacy jobs (before two-pass was added)
-			pass1Job = j
-		}
-		if pass1Job != nil && pass2Job != nil {
-			break
+			if pass1Meta == nil || m.StartedAt.After(pass1Meta.StartedAt) {
+				pass1Meta = m
+			}
 		}
 	}
 
-	if pass1Job == nil && pass2Job == nil {
+	if pass1Meta == nil && pass2Meta == nil {
 		return nil, nil, fmt.Errorf("%w: %s", errNoScans, serverName)
 	}
 
+	var pass1Job, pass2Job *ScanJob
+	if pass1Meta != nil {
+		if pass1Job, err = s.storage.GetScanJob(pass1Meta.ID); err != nil {
+			return nil, nil, fmt.Errorf("load latest pass-1 job %s: %w", pass1Meta.ID, err)
+		}
+	}
+	if pass2Meta != nil {
+		if pass2Job, err = s.storage.GetScanJob(pass2Meta.ID); err != nil {
+			return nil, nil, fmt.Errorf("load latest pass-2 job %s: %w", pass2Meta.ID, err)
+		}
+	}
+
 	return pass1Job, pass2Job, nil
 }
 

internal/security/scanner/service_test.go

@@ -93,6 +93,26 @@ func (m *mockStorage) ListScanJobs(serverName string) ([]*ScanJob, error) {
 	return result, nil
 }
 
+func (m *mockStorage) ListScanJobMetas(serverName string) ([]*ScanJobMeta, error) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	var result []*ScanJobMeta
+	for _, j := range m.jobs {
+		if serverName != "" && j.ServerName != serverName {
+			continue
+		}
+		result = append(result, &ScanJobMeta{
+			ID:          j.ID,
+			ServerName:  j.ServerName,
+			Status:      j.Status,
+			ScanPass:    j.ScanPass,
+			StartedAt:   j.StartedAt,
+			CompletedAt: j.CompletedAt,
+		})
+	}
+	return result, nil
+}
+
 func (m *mockStorage) GetLatestScanJob(serverName string) (*ScanJob, error) {
 	m.mu.Lock()
 	defer m.mu.Unlock()
@@ -1545,12 +1565,14 @@ func TestScanFindingScanPassTag(t *testing.T) {
 	}
 }
 
-// countingStorage wraps any Storage and counts ListScanJobs invocations.
-// Used for verifying the GetScanSummary negative-cache behavior introduced in
-// spec 047.
+// countingStorage wraps any Storage and counts storage probes. listCalls counts
+// the heavy full-history ListScanJobs path (MCP-2205 asserts this stays 0 on the
+// report hot paths); metaCalls counts the lightweight ListScanJobMetas index
+// path used by the GetScanSummary negative-cache behavior (spec 047).
 type countingStorage struct {
 	Storage
 	listCalls atomic.Int64
+	metaCalls atomic.Int64
 }
 
 func newCountingStorage(inner Storage) *countingStorage {
@@ -1562,6 +1584,11 @@ func (c *countingStorage) ListScanJobs(serverName string) ([]*ScanJob, error) {
 	return c.Storage.ListScanJobs(serverName)
 }
 
+func (c *countingStorage) ListScanJobMetas(serverName string) ([]*ScanJobMeta, error) {
+	c.metaCalls.Add(1)
+	return c.Storage.ListScanJobMetas(serverName)
+}
+
 // erroringStorage returns a transient error from ListScanJobs while delegating
 // other calls to the inner Storage. Used to verify that non-errNoScans errors
 // do NOT populate the negative cache.
@@ -1576,6 +1603,11 @@ func (e *erroringStorage) ListScanJobs(string) ([]*ScanJob, error) {
 	return nil, e.err
 }
 
+func (e *erroringStorage) ListScanJobMetas(string) ([]*ScanJobMeta, error) {
+	e.listCalls.Add(1)
+	return nil, e.err
+}
+
 // Spec 047 — Phase 3 (US1): cache the "no scans found" sentinel.
 
 func TestGetScanSummary_CachesNegativeResult(t *testing.T) {
@@ -1590,8 +1622,8 @@ func TestGetScanSummary_CachesNegativeResult(t *testing.T) {
 	}
 
 	// Without the negative-cache fix, this would be N storage calls.
-	if got := store.listCalls.Load(); got != 1 {
-		t.Errorf("expected exactly 1 ListScanJobs call after %d GetScanSummary invocations, got %d", N, got)
+	if got := store.metaCalls.Load(); got != 1 {
+		t.Errorf("expected exactly 1 ListScanJobMetas call after %d GetScanSummary invocations, got %d", N, got)
 	}
 }
 
@@ -1606,7 +1638,7 @@ func TestGetScanSummary_DoesNotCacheOnTransientError(t *testing.T) {
 
 	// Transient error must NOT populate the negative cache: every call retries.
 	if got := store.listCalls.Load(); got != int64(N) {
-		t.Errorf("expected %d ListScanJobs calls (no caching of transient errors), got %d", N, got)
+		t.Errorf("expected %d storage probes (no caching of transient errors), got %d", N, got)
 	}
 }
 
@@ -1619,8 +1651,8 @@ func TestGetScanSummary_OverwritesNilSentinelOnRealScan(t *testing.T) {
 	if got := svc.GetScanSummary(context.Background(), "later-scanned"); got != nil {
 		t.Fatalf("expected nil summary on first call, got %+v", got)
 	}
-	if got := store.listCalls.Load(); got != 1 {
-		t.Fatalf("expected 1 ListScanJobs call after first GetScanSummary, got %d", got)
+	if got := store.metaCalls.Load(); got != 1 {
+		t.Fatalf("expected 1 ListScanJobMetas call after first GetScanSummary, got %d", got)
 	}
 
 	// Simulate a real scan landing for that server: insert a completed Pass-1 job
@@ -1646,3 +1678,95 @@ func TestGetScanSummary_OverwritesNilSentinelOnRealScan(t *testing.T) {
 		t.Errorf("expected real summary {Status: clean}, got %+v", got)
 	}
 }
+
+// TestGetScanReportByJobID_LatencyIndependentOfHistory verifies the MCP-2205
+// fix: aggregating a Pass-1 report must NOT deserialize the full per-server scan
+// history (ListScanJobs), whose job payloads carry large stdout/stderr. The
+// companion Pass-2 lookup uses the lightweight metadata index instead, so report
+// latency does not grow with how many times a server has been scanned.
+func TestGetScanReportByJobID_LatencyIndependentOfHistory(t *testing.T) {
+	mock := newMockStorage()
+	store := newCountingStorage(mock)
+	svc := NewService(store, NewRegistry(t.TempDir(), zap.NewNop()), nil, t.TempDir(), zap.NewNop())
+
+	now := time.Now()
+
+	// Pass-1 job under inspection + its report.
+	pass1 := &ScanJob{ID: "p1", ServerName: "srv", Status: ScanJobStatusCompleted, ScanPass: ScanPassSecurityScan, StartedAt: now.Add(-10 * time.Minute)}
+	_ = mock.SaveScanJob(pass1)
+	_ = mock.SaveScanReport(&ScanReport{ID: "r1", JobID: "p1", ServerName: "srv", ScannerID: "s",
+		Findings: []ScanFinding{{RuleID: "T1", Title: "tool poisoning", Scanner: "s", ThreatLevel: ThreatLevelDangerous}}})
+
+	// Companion Pass-2 job that started after Pass-1 + its report.
+	pass2 := &ScanJob{ID: "p2", ServerName: "srv", Status: ScanJobStatusCompleted, ScanPass: ScanPassSupplyChainAudit, StartedAt: now.Add(-8 * time.Minute)}
+	_ = mock.SaveScanJob(pass2)
+	_ = mock.SaveScanReport(&ScanReport{ID: "r2", JobID: "p2", ServerName: "srv", ScannerID: "s",
+		Findings: []ScanFinding{{RuleID: "CVE-1", Title: "known cve", Scanner: "s", ThreatLevel: ThreatLevelWarning}}})
+
+	// Large historical backlog for the same server (the symptom: reports get
+	// slower the more a server is scanned).
+	for i := 0; i < 100; i++ {
+		_ = mock.SaveScanJob(&ScanJob{
+			ID:         fmt.Sprintf("noise-%d", i),
+			ServerName: "srv",
+			Status:     ScanJobStatusCompleted,
+			ScanPass:   ScanPassSecurityScan,
+			StartedAt:  now.Add(time.Duration(-20-i) * time.Minute),
+		})
+	}
+
+	agg, err := svc.GetScanReportByJobID(context.Background(), "p1")
+	if err != nil {
+		t.Fatalf("GetScanReportByJobID: %v", err)
+	}
+
+	// Correctness: companion Pass-2 still merged in.
+	if !agg.Pass1Complete || !agg.Pass2Complete {
+		t.Errorf("expected both passes complete, got pass1=%v pass2=%v", agg.Pass1Complete, agg.Pass2Complete)
+	}
+	if len(agg.Findings) != 2 {
+		t.Fatalf("expected 2 merged findings (Pass-1 + companion Pass-2), got %d", len(agg.Findings))
+	}
+
+	// Latency guard: must not scan the full per-server job history.
+	if got := store.listCalls.Load(); got != 0 {
+		t.Errorf("expected 0 ListScanJobs (full-history) calls in report path, got %d", got)
+	}
+}
+
+// TestGetScanReport_LatestLatencyIndependentOfHistory verifies the MCP-2205 fix
+// also covers the "latest report" path (GetScanReport by server name, used by the
+// Web UI server-detail view). Resolving the latest Pass-1/Pass-2 jobs must use the
+// lightweight metadata index plus targeted GetScanJob loads (at most two full job
+// deserializations), not a full ListScanJobs over the server's scan history.
+func TestGetScanReport_LatestLatencyIndependentOfHistory(t *testing.T) {
+	mock := newMockStorage()
+	store := newCountingStorage(mock)
+	svc := NewService(store, NewRegistry(t.TempDir(), zap.NewNop()), nil, t.TempDir(), zap.NewNop())
+
+	now := time.Now()
+
+	// Latest Pass-1 + Pass-2 with reports.
+	_ = mock.SaveScanJob(&ScanJob{ID: "latest-p1", ServerName: "srv", Status: ScanJobStatusCompleted, ScanPass: ScanPassSecurityScan, StartedAt: now.Add(-2 * time.Minute)})
+	_ = mock.SaveScanReport(&ScanReport{ID: "lr1", JobID: "latest-p1", ServerName: "srv", ScannerID: "s",
+		Findings: []ScanFinding{{RuleID: "T1", Title: "tp", Scanner: "s", ThreatLevel: ThreatLevelDangerous}}})
+	_ = mock.SaveScanJob(&ScanJob{ID: "latest-p2", ServerName: "srv", Status: ScanJobStatusCompleted, ScanPass: ScanPassSupplyChainAudit, StartedAt: now.Add(-1 * time.Minute)})
+	_ = mock.SaveScanReport(&ScanReport{ID: "lr2", JobID: "latest-p2", ServerName: "srv", ScannerID: "s",
+		Findings: []ScanFinding{{RuleID: "CVE-9", Title: "cve", Scanner: "s", ThreatLevel: ThreatLevelWarning}}})
+
+	// Large historical backlog.
+	for i := 0; i < 100; i++ {
+		_ = mock.SaveScanJob(&ScanJob{ID: fmt.Sprintf("old-%d", i), ServerName: "srv", Status: ScanJobStatusCompleted, ScanPass: ScanPassSecurityScan, StartedAt: now.Add(time.Duration(-10-i) * time.Minute)})
+	}
+
+	report, err := svc.GetScanReport(context.Background(), "srv")
+	if err != nil {
+		t.Fatalf("GetScanReport: %v", err)
+	}
+	if len(report.Findings) != 2 {
+		t.Fatalf("expected 2 merged findings, got %d", len(report.Findings))
+	}
+	if got := store.listCalls.Load(); got != 0 {
+		t.Errorf("expected 0 ListScanJobs (full-history) calls in latest-report path, got %d", got)
+	}
+}

internal/security/scanner/types.go

@@ -120,6 +120,20 @@ type ScanJob struct {
 	ScanContext *ScanContext `json:"scan_context,omitempty"`
 }
 
+// ScanJobMeta is a lightweight projection of a scan job, persisted in a
+// dedicated index bucket so that companion-job lookups during report
+// aggregation never deserialize the full job payload (whose ScannerStatuses can
+// carry large stdout/stderr blobs). This keeps report latency independent of a
+// server's scan history. See MCP-2205.
+type ScanJobMeta struct {
+	ID          string    `json:"id"`
+	ServerName  string    `json:"server_name"`
+	Status      string    `json:"status"`
+	ScanPass    int       `json:"scan_pass"`
+	StartedAt   time.Time `json:"started_at"`
+	CompletedAt time.Time `json:"completed_at,omitempty"`
+}
+
 // ScanJobSummary is a lightweight view of a scan job for history listing
 type ScanJobSummary struct {
 	ID            string    `json:"id"`

internal/storage/bbolt.go

@@ -93,6 +93,7 @@ func (b *BoltDB) initBuckets() error {
 			ActivityStatsBucket,
 			ScannersBucket,
 			ScanJobsBucket,
+			ScanJobIndexBucket,
 			ScanReportsBucket,
 			IntegrityBaselinesBucket,
 			OnboardingBucket,
@@ -104,6 +105,12 @@ func (b *BoltDB) initBuckets() error {
 			}
 		}
 
+		// Backfill the scan-job index for databases created before MCP-2205.
+		// Idempotent: only runs when the index is empty but jobs exist.
+		if err := backfillScanJobIndex(tx); err != nil {
+			return fmt.Errorf("failed to backfill scan job index: %w", err)
+		}
+
 		// Set schema version only for new databases. Existing databases keep their
 		// stored version so migrations can observe and upgrade them.
 		metaBucket := tx.Bucket([]byte(MetaBucket))