feat(httpapi): harden control plane and speed up prefix refresh

Очередь job получила lease и fencing, телеметрия и прогресс deploy сохраняются, ревизия пишет manifest входов. Обновление префиксов не переписывает неизменённый снимок, большие наборы идут через COPY, а устаревшая ревизия не попадает в deploy. Бандл ноды кэшируется, prune и постановка job больше не держат горячий путь.

Co-authored-by: Cursor <cursoragent@cursor.com>

internal/store/prefix_snapshot_reuse_test.go

@@ -0,0 +1,153 @@
+package store
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+)
+
+func TestSameSnapshotContentIgnoresOrder(t *testing.T) {
+	comm := "c1"
+	a := &ModulePrefixSnapshot{InputHash: "h", Prefixes: []PrefixRow{
+		{Prefix: "203.0.113.0/24", CommunityID: &comm, Source: "cdn:a"},
+		{Prefix: "198.51.100.0/24", Source: "cdn:b"},
+	}}
+	b := &ModulePrefixSnapshot{InputHash: "h", Prefixes: []PrefixRow{
+		{Prefix: "198.51.100.0/24", Source: "cdn:b"},
+		{Prefix: "203.0.113.0/24", CommunityID: &comm, Source: "cdn:a"},
+	}}
+	if !SameSnapshotContent(a, b) {
+		t.Fatal("same set in another order was treated as new")
+	}
+	b.Prefixes[0].Prefix = "192.0.2.0/24"
+	if SameSnapshotContent(a, b) {
+		t.Fatal("different prefix compared equal")
+	}
+}
+
+func TestPrefixSnapshotHashIgnoresOrder(t *testing.T) {
+	rows := []PrefixRow{
+		{Prefix: "203.0.113.0/24", Source: "ip_range"},
+		{Prefix: "198.51.100.0/24", Source: "cdn:a"},
+	}
+	rev := []PrefixRow{rows[1], rows[0]}
+	if PrefixSnapshotHash(rows) != PrefixSnapshotHash(rev) {
+		t.Fatal("prefix hash depends on order")
+	}
+	rev = append(rev, PrefixRow{Prefix: "2001:db8::/32", Source: "ip_range"})
+	if PrefixSnapshotHash(rows) == PrefixSnapshotHash(rev) {
+		t.Fatal("prefix change kept the same hash")
+	}
+}
+
+func TestCreateRenderRevisionReusesPrefixSnapshot(t *testing.T) {
+	m := NewMemory()
+	m.SeedDemo()
+	tenant, _, mod, _, _ := m.DemoIDs()
+	ctx := context.Background()
+	rows := []PrefixRow{
+		{Prefix: "203.0.113.0/24", Source: "ip_range"},
+		{Prefix: "198.51.100.0/24", Source: "cdn:a"},
+	}
+	rev1 := uuid.NewString()
+	if err := m.CreateRenderRevision(ctx, rev1, tenant, mod, nil, "hash-peer-a", map[string]string{"bird.conf": "a"}, rows); err != nil {
+		t.Fatal(err)
+	}
+	rev2 := uuid.NewString()
+	reversed := []PrefixRow{rows[1], rows[0]}
+	if err := m.CreateRenderRevision(ctx, rev2, tenant, mod, nil, "hash-peer-b", map[string]string{"bird.conf": "b"}, reversed); err != nil {
+		t.Fatal(err)
+	}
+	if m.revPrefixSnap[rev1] == "" || m.revPrefixSnap[rev1] != m.revPrefixSnap[rev2] {
+		t.Fatalf("peer-only change copied prefix rows: %q vs %q", m.revPrefixSnap[rev1], m.revPrefixSnap[rev2])
+	}
+	if len(m.prefixSnapRows) != 1 {
+		t.Fatalf("stored prefix snapshots: %d", len(m.prefixSnapRows))
+	}
+	page, _, _ := m.ListRevisionPrefixes(tenant, rev2, "", 10)
+	if len(page) != 2 {
+		t.Fatalf("reused snapshot not readable: %+v", page)
+	}
+	changed := append(append([]PrefixRow(nil), rows...), PrefixRow{Prefix: "2001:db8::/32", Source: "ip_range"})
+	rev3 := uuid.NewString()
+	if err := m.CreateRenderRevision(ctx, rev3, tenant, mod, nil, "hash-prefix", nil, changed); err != nil {
+		t.Fatal(err)
+	}
+	if m.revPrefixSnap[rev3] == m.revPrefixSnap[rev1] || len(m.prefixSnapRows) != 2 {
+		t.Fatalf("prefix change did not allocate a snapshot: id=%s count=%d", m.revPrefixSnap[rev3], len(m.prefixSnapRows))
+	}
+}
+
+func TestUpdateModulePrefixSnapshotCancelledContext(t *testing.T) {
+	m := NewMemory()
+	m.SeedDemo()
+	tenant, mod, _, _, _ := m.DemoIDs()
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	err := m.UpdateModulePrefixSnapshot(ctx, tenant, mod, func(*ModulePrefixSnapshot) (*ModulePrefixSnapshot, error) {
+		t.Fatal("snapshot update ran after cancel")
+		return nil, nil
+	})
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("err=%v", err)
+	}
+}
+
+func TestMemorySnapshotLockUsesSnapshotTimeout(t *testing.T) {
+	t.Setenv("EVOBGP_SNAPSHOT_TIMEOUT", "80ms")
+	m := NewMemory()
+	hold, err := m.lockModuleSnapshot(context.Background(), "tenant", "module")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer hold()
+	start := time.Now()
+	_, err = m.lockModuleSnapshot(context.Background(), "tenant", "module")
+	if !errors.Is(err, context.DeadlineExceeded) {
+		t.Fatalf("err=%v", err)
+	}
+	if time.Since(start) > time.Second {
+		t.Fatalf("lock wait ignored snapshot timeout: %s", time.Since(start))
+	}
+}
+
+func TestSnapshotWriteBudgetKeepsCallerDeadline(t *testing.T) {
+	parent, cancel := context.WithTimeout(context.Background(), time.Hour)
+	defer cancel()
+	got, stop, installed := SnapshotWriteBudget(parent)
+	defer stop()
+	if installed {
+		t.Fatal("caller deadline was replaced")
+	}
+	dl, ok := got.Deadline()
+	if !ok || time.Until(dl) < 30*time.Minute {
+		t.Fatal("caller deadline was shortened")
+	}
+}
+
+func TestSetASNPrefixCacheKeepsRowsWhenSetUnchanged(t *testing.T) {
+	m := NewMemory()
+	if err := m.SetASNPrefixCache(64500, "holder-a", []string{"203.0.113.0/24", "198.51.100.0/24"}); err != nil {
+		t.Fatal(err)
+	}
+	before := m.asnPrefixCache[64500]
+	ptr := &before.Prefixes[0]
+	fetched := before.FetchedAt
+	time.Sleep(5 * time.Millisecond)
+	if err := m.SetASNPrefixCache(64500, "holder-b", []string{"198.51.100.0/24", "203.0.113.0/24"}); err != nil {
+		t.Fatal(err)
+	}
+	after := m.asnPrefixCache[64500]
+	if &after.Prefixes[0] != ptr {
+		t.Fatal("unchanged ASN prefixes were rewritten")
+	}
+	if !after.FetchedAt.After(fetched) {
+		t.Fatal("fetched_at was not updated")
+	}
+	if after.Holder != "holder-b" {
+		t.Fatalf("holder=%q", after.Holder)
+	}
+}