| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "fmt" |
| 6 | "testing" |
| 7 | "time" |
| 8 | |
| 9 | "github.com/carlosframework/platform/internal/hostbox" |
| 10 | "github.com/carlosframework/platform/internal/hostclaim" |
| 11 | "github.com/carlosframework/platform/internal/hostguard" |
| 12 | "github.com/carlosframework/platform/internal/provision" |
| 13 | ) |
| 14 | |
| 15 | func plan14InstallRegistration(t *testing.T, kind, phase string) (*recoveryFixture, ClaimRootHTTP, hostbox.JournalRequest) { |
| 16 | t.Helper() |
| 17 | f := newRecoveryFixture(t) |
| 18 | if _, err := hostclaim.Update(t.Context(), f.claims, f.read(), func(c *hostclaim.Claim) error { |
| 19 | if kind != hostclaim.KindServing { |
| 20 | c.Op = hostclaim.Op{ID: "install-op", Kind: kind, Phase: phase, Exec: "box-1", Participants: []string{"box-0", "box-1"}, Intent: map[string]string{"source": "box-0", "target": "box-1", "to": "ie", "region": "ie", "source_box": "box-0", "target_box": "box-1"}, Journal: map[string]string{"install_stamp": fmt.Sprint(c.Holder.Stamp)}} |
| 21 | } |
| 22 | return nil |
| 23 | }); err != nil { |
| 24 | t.Fatal(err) |
| 25 | } |
| 26 | o := f.read() |
| 27 | h := o.Claim.Holder |
| 28 | rec := provision.Record{Schema: 1, Host: o.Host, Account: h.Account, App: h.App, Channel: "stable", Region: "ie", CreatedAt: f.now} |
| 29 | if kind == hostclaim.KindLanding { |
| 30 | rec.Move = &provision.MoveState{ID: o.Claim.Op.ID, Incarnation: h.Incarnation, From: "uk", To: "ie", State: provision.MoveLanding, LandingPhase: phase} |
| 31 | } |
| 32 | plan14PutRecord(t, f.claims, rec) |
| 33 | _, etag, err := f.claims.Get(t.Context(), h.Key) |
| 34 | if err != nil { |
| 35 | t.Fatal(err) |
| 36 | } |
| 37 | reg := restartTestRegistry(t) |
| 38 | journals, err := hostbox.NewJournalService(f.journal) |
| 39 | if err != nil { |
| 40 | t.Fatal(err) |
| 41 | } |
| 42 | guard := f.guard |
| 43 | guard.Root = hostguard.NewRootAuthority(&rollbackDisk{dir: t.TempDir()}) |
| 44 | actor := hostbox.Actor{Box: "box-1", Enrollment: "en", Region: "ie", Revision: 1, Binding: "1"} |
| 45 | roots := map[string]string{"data": guard.DataDir, "opt": t.TempDir(), "units": t.TempDir()} |
| 46 | census := &hostbox.EmptyInstallationCensus{Inventory: journals.Inventory, Registry: reg, Launches: f.journal, Box: actor.Box, Enrollment: actor.Enrollment, DirectoryRevision: "1", Roots: roots, SharedPaths: map[string][]string{"data": {"hostguard"}}} |
| 47 | root := ClaimRootHTTP{Journals: journals, Launches: f.journal, Roots: roots, InventoryActor: func(context.Context) (hostbox.Actor, error) { return actor, nil }, Reconcile: hostbox.ReconcileAdapter{Registry: reg, Inventory: journals.Inventory, Census: census}, Stop: ClaimStopHTTP{Coordinator: ClaimStopCoordinator{Guard: guard, Evidence: f.claims}}} |
| 48 | q := hostbox.JournalRequest{V: 3, ID: "registration-1", Op: "inventory-write", RecordKey: h.Key, RecordETag: etag, Target: hostguard.FenceTarget{Host: o.Host, Incarnation: h.Incarnation, Stamp: h.Stamp, Session: h.Session, Box: actor.Box, Enrollment: actor.Enrollment, ClaimETag: o.ETag, Op: o.Claim.Op.ID, Phase: o.Claim.Op.Phase}, Inventory: hostbox.StopInventory{Complete: true, Host: o.Host, Incarnation: h.Incarnation, InstallationStamp: h.Stamp, Box: actor.Box, Enrollment: actor.Enrollment, DirectoryRevision: "1"}} |
| 49 | _, client := hostbox.RemoteJournals(func(_ context.Context, prepared hostbox.JournalRequest) (hostbox.JournalResponse, error) { |
| 50 | q = prepared |
| 51 | return hostbox.JournalResponse{V: 1, ID: prepared.ID, Revision: 1}, nil |
| 52 | }) |
| 53 | if _, err := client.WriteExpected(t.Context(), q.Inventory, hostbox.InventoryExpectation{RecordKey: q.RecordKey, RecordETag: q.RecordETag, Target: q.Target}); err != nil { |
| 54 | t.Fatal(err) |
| 55 | } |
| 56 | return f, root, q |
| 57 | } |
| 58 | |
| 59 | func TestPlan14FirstRegistrationOperations(t *testing.T) { |
| 60 | t.Parallel() |
| 61 | for _, kind := range []string{hostclaim.KindServing, hostclaim.KindRelocating, hostclaim.KindLanding} { |
| 62 | for _, attack := range []string{"none", "phase", "box", "session", "enrollment", "record", "op", "stamp", "agent row"} { |
| 63 | t.Run(kind+"/"+attack, func(t *testing.T) { |
| 64 | phase := hostclaim.PhaseInstalling |
| 65 | if kind == hostclaim.KindLanding { |
| 66 | phase = provision.LandChecking |
| 67 | } |
| 68 | f, root, q := plan14InstallRegistration(t, kind, phase) |
| 69 | switch attack { |
| 70 | case "phase": |
| 71 | q.Target.Phase = "wrong" |
| 72 | case "box": |
| 73 | q.Target.Box = "box-0" |
| 74 | case "session": |
| 75 | q.Target.Session++ |
| 76 | case "enrollment": |
| 77 | q.Target.Enrollment = "other" |
| 78 | case "record": |
| 79 | q.RecordETag = "other" |
| 80 | case "op": |
| 81 | q.Target.Op = "other" |
| 82 | case "stamp": |
| 83 | q.Target.Stamp++ |
| 84 | case "agent row": |
| 85 | rt, _, err := rootClaimRecord(t.Context(), f.claims, f.read().Claim, root.Roots["data"]) |
| 86 | if err != nil { |
| 87 | t.Fatal(err) |
| 88 | } |
| 89 | if err = root.Reconcile.Registry.Upsert(rt); err != nil { |
| 90 | t.Fatal(err) |
| 91 | } |
| 92 | } |
| 93 | out, err := root.commitRootInventory(t.Context(), q) |
| 94 | if attack != "none" { |
| 95 | if err == nil { |
| 96 | t.Fatal("unauthorized registration") |
| 97 | } |
| 98 | return |
| 99 | } |
| 100 | if err != nil { |
| 101 | t.Fatal(err) |
| 102 | } |
| 103 | if out.Revision != 1 { |
| 104 | t.Fatal(out) |
| 105 | } |
| 106 | if _, err = root.commitRootInventory(t.Context(), q); err != nil { |
| 107 | t.Fatal("replay", err) |
| 108 | } |
| 109 | if kind == hostclaim.KindLanding { |
| 110 | row, err := root.Reconcile.Registry.Get(q.Target.Host) |
| 111 | if err != nil || row == nil || row.HoldOp != q.Target.Op || row.Status != "landing" || row.DBPath != provision.DefaultDBPath(root.Roots["data"], q.Target.Host) || !row.ClaimOwnership.Valid() { |
| 112 | t.Fatalf("root held row: %+v %v", row, err) |
| 113 | } |
| 114 | } |
| 115 | }) |
| 116 | } |
| 117 | } |
| 118 | } |
| 119 | |
| 120 | func TestPlan14FirstRegistrationRefusesLaterPhases(t *testing.T) { |
| 121 | t.Parallel() |
| 122 | for kind, phases := range map[string][]string{hostclaim.KindLanding: {provision.LandConfigured, provision.LandRestored, provision.LandProbing}, hostclaim.KindRelocating: {hostclaim.PhaseRelocatingProbing, hostclaim.PhaseProbeFailed}, hostclaim.KindRestoring: {provision.LandConfigured}, hostclaim.KindReverting: {provision.LandConfigured}} { |
| 123 | for _, phase := range phases { |
| 124 | t.Run(kind+"/"+phase, func(t *testing.T) { |
| 125 | _, root, q := plan14InstallRegistration(t, kind, phase) |
| 126 | if _, err := root.commitRootInventory(t.Context(), q); err == nil { |
| 127 | t.Fatal("skipped first registration") |
| 128 | } |
| 129 | }) |
| 130 | } |
| 131 | } |
| 132 | } |
| 133 | |
| 134 | func plan14PreparedInventory(t *testing.T, root ClaimRootHTTP, q hostbox.JournalRequest) hostbox.JournalRequest { |
| 135 | t.Helper() |
| 136 | _, client := hostbox.RemoteJournals(func(_ context.Context, p hostbox.JournalRequest) (hostbox.JournalResponse, error) { |
| 137 | q = p |
| 138 | return hostbox.JournalResponse{V: 1, ID: p.ID, Revision: p.Revision + 1}, nil |
| 139 | }) |
| 140 | if _, err := client.WriteExpected(t.Context(), q.Inventory, hostbox.InventoryExpectation{Revision: q.Revision, RecordKey: q.RecordKey, RecordETag: q.RecordETag, Target: q.Target}); err != nil { |
| 141 | t.Fatal(err) |
| 142 | } |
| 143 | return q |
| 144 | } |
| 145 | |
| 146 | func TestPlan14LandingConfiguredAdoptsOnlyRootRow(t *testing.T) { |
| 147 | t.Parallel() |
| 148 | for _, attack := range []string{"none", "row id", "row missing", "op", "session", "record", "enrollment", "probe session"} { |
| 149 | t.Run(attack, func(t *testing.T) { |
| 150 | f, root, q := plan14InstallRegistration(t, hostclaim.KindLanding, provision.LandChecking) |
| 151 | if _, err := root.commitRootInventory(t.Context(), q); err != nil { |
| 152 | t.Fatal(err) |
| 153 | } |
| 154 | prior, err := root.Reconcile.Registry.Get(q.Target.Host) |
| 155 | if err != nil { |
| 156 | t.Fatal(err) |
| 157 | } |
| 158 | o, err := hostclaim.Update(t.Context(), f.claims, f.read(), func(c *hostclaim.Claim) error { |
| 159 | c.Op.Phase = provision.LandConfigured |
| 160 | if attack == "probe session" { |
| 161 | c.Op.Phase = provision.LandProbing |
| 162 | } |
| 163 | if attack == "op" { |
| 164 | c.Op.ID = "different" |
| 165 | } |
| 166 | if attack == "session" { |
| 167 | c.Holder.Session++ |
| 168 | } |
| 169 | return nil |
| 170 | }) |
| 171 | if err != nil { |
| 172 | t.Fatal(err) |
| 173 | } |
| 174 | rec, _, err := provision.FindRecordByHost(t.Context(), f.claims, o.Host) |
| 175 | if err != nil { |
| 176 | t.Fatal(err) |
| 177 | } |
| 178 | if attack == "probe session" { |
| 179 | h := o.Claim.Holder |
| 180 | o, err = hostclaim.Update(t.Context(), f.claims, o, hostclaim.AdmitWriter(hostclaim.Admission{ObservedOpID: o.Claim.Op.ID, ObservedPhase: o.Claim.Op.Phase, ObservedSession: h.Session, ObservedGeneration: h.Generation, Box: "box-1", Mode: "cold", NewGeneration: "landing-generation", KeepStamp: true, Coverage: "sealed", Region: "ie", PredecessorRegion: "uk"})) |
| 181 | if err != nil { |
| 182 | t.Fatal(err) |
| 183 | } |
| 184 | } |
| 185 | rec.Move.LandingPhase = o.Claim.Op.Phase |
| 186 | rec.Move.ID = o.Claim.Op.ID |
| 187 | plan14PutRecord(t, f.claims, *rec) |
| 188 | _, q.RecordETag, err = f.claims.Get(t.Context(), q.RecordKey) |
| 189 | if err != nil { |
| 190 | t.Fatal(err) |
| 191 | } |
| 192 | q.Target.ClaimETag, q.Target.Phase, q.Target.Op, q.Target.Session = o.ETag, o.Claim.Op.Phase, o.Claim.Op.ID, o.Claim.Holder.Session |
| 193 | q.Inventory, q.Revision, err = root.Journals.Inventory.Snapshot(t.Context(), o.Host, hostguard.Captured{Incarnation: q.Target.Incarnation, Stamp: q.Target.Stamp}) |
| 194 | if err != nil { |
| 195 | t.Fatal(err) |
| 196 | } |
| 197 | switch attack { |
| 198 | case "row id": |
| 199 | prior.ClaimRowID = "forged" |
| 200 | if err = root.Reconcile.Registry.Delete(o.Host); err == nil { |
| 201 | err = root.Reconcile.Registry.Upsert(*prior) |
| 202 | } |
| 203 | case "row missing": |
| 204 | err = root.Reconcile.Registry.Delete(o.Host) |
| 205 | case "record": |
| 206 | q.RecordETag = "stale" |
| 207 | case "enrollment": |
| 208 | q.Target.Enrollment = "stale" |
| 209 | } |
| 210 | if err != nil { |
| 211 | t.Fatal(err) |
| 212 | } |
| 213 | q = plan14PreparedInventory(t, root, q) |
| 214 | out, err := root.commitRootInventory(t.Context(), q) |
| 215 | if attack != "none" && attack != "probe session" { |
| 216 | if err == nil { |
| 217 | t.Fatal("adopted changed installation") |
| 218 | } |
| 219 | return |
| 220 | } |
| 221 | if err != nil { |
| 222 | t.Fatal(err) |
| 223 | } |
| 224 | if out.Revision != 2 { |
| 225 | t.Fatal(out) |
| 226 | } |
| 227 | row, err := root.Reconcile.Registry.Get(o.Host) |
| 228 | if err != nil || row.ClaimOwnership != prior.ClaimOwnership { |
| 229 | t.Fatalf("identity changed: %+v %v", row, err) |
| 230 | } |
| 231 | if _, err := root.commitRootInventory(t.Context(), q); err != nil { |
| 232 | t.Fatal("configured replay", err) |
| 233 | } |
| 234 | }) |
| 235 | } |
| 236 | } |
| 237 | |
| 238 | func TestPlan14LandingRegistrationOutsideHostLock(t *testing.T) { |
| 239 | t.Parallel() |
| 240 | for _, stale := range []bool{false, true} { |
| 241 | t.Run(fmt.Sprint(stale), func(t *testing.T) { |
| 242 | f, root, _ := plan14InstallRegistration(t, hostclaim.KindLanding, provision.LandChecking) |
| 243 | if err := f.guard.Activate(t.Context(), f.read().Host, "box-1", nil); err != nil { |
| 244 | t.Fatal(err) |
| 245 | } |
| 246 | previous, err := f.guard.Capture(f.read().Host) |
| 247 | if err != nil { |
| 248 | t.Fatal(err) |
| 249 | } |
| 250 | if _, err := hostclaim.Update(t.Context(), f.claims, f.read(), func(c *hostclaim.Claim) error { c.Holder.Stamp++; return nil }); err != nil { |
| 251 | t.Fatal(err) |
| 252 | } |
| 253 | guard := f.guard |
| 254 | guard.FenceView = root.Stop.Coordinator.Guard.Root.View |
| 255 | _, remote := hostbox.RemoteJournals(func(ctx context.Context, q hostbox.JournalRequest) (hostbox.JournalResponse, error) { |
| 256 | if q.Op != "inventory-write" { |
| 257 | return root.Journals.Execute(ctx, "box-1/en", q) |
| 258 | } |
| 259 | out, err := root.commitRootInventory(ctx, q) |
| 260 | if err == nil && stale { |
| 261 | rec, _, e := provision.FindRecordByHost(ctx, f.claims, q.Target.Host) |
| 262 | if e != nil { |
| 263 | return out, e |
| 264 | } |
| 265 | rec.RecordSeq++ |
| 266 | plan14PutRecord(t, f.claims, *rec) |
| 267 | } |
| 268 | return out, err |
| 269 | }) |
| 270 | a := hostbox.ReconcileAdapter{Inventory: remote, Box: "box-1", Enrollment: "en", DirectoryRevision: "1", Reconciler: hostbox.Reconciler{Claims: f.claims, Guard: guard}} |
| 271 | ctx, cancel := context.WithTimeout(t.Context(), 2*time.Second) |
| 272 | defer cancel() |
| 273 | err = a.RegisterLanding(ctx, f.read()) |
| 274 | if stale { |
| 275 | if err == nil { |
| 276 | t.Fatal("activated after record changed during RPC") |
| 277 | } |
| 278 | return |
| 279 | } |
| 280 | if err != nil { |
| 281 | t.Fatal("registration blocked or refused", err) |
| 282 | } |
| 283 | if err := a.RegisterLanding(ctx, f.read()); err != nil { |
| 284 | t.Fatal("checking retry", err) |
| 285 | } |
| 286 | captured, err := f.guard.Capture(f.read().Host) |
| 287 | if err != nil || captured != previous { |
| 288 | t.Fatalf("registration hid previous stay: %+v, want %+v: %v", captured, previous, err) |
| 289 | } |
| 290 | if err := guard.Activate(ctx, f.read().Host, "box-1", nil); err != nil { |
| 291 | t.Fatal("ordered activation", err) |
| 292 | } |
| 293 | }) |
| 294 | } |
| 295 | } |
| 296 | |
| 297 | type plan14RegistrationCrash struct { |
| 298 | *rollbackDisk |
| 299 | writes int |
| 300 | } |
| 301 | |
| 302 | func (s *plan14RegistrationCrash) Write(name string, body []byte) error { |
| 303 | s.writes++ |
| 304 | if s.writes == 2 { |
| 305 | return fmt.Errorf("injected root commit crash") |
| 306 | } |
| 307 | return s.rollbackDisk.Write(name, body) |
| 308 | } |
| 309 | |
| 310 | func TestPlan14LandingRegistrationRecoversRootCommit(t *testing.T) { |
| 311 | _, root, q := plan14InstallRegistration(t, hostclaim.KindLanding, provision.LandChecking) |
| 312 | store := &plan14RegistrationCrash{rollbackDisk: &rollbackDisk{dir: t.TempDir()}} |
| 313 | root.Stop.Coordinator.Guard.Root = hostguard.NewRootAuthority(store) |
| 314 | if _, err := root.commitRootInventory(t.Context(), q); err == nil { |
| 315 | t.Fatal("missed injected crash") |
| 316 | } |
| 317 | before, err := root.Reconcile.Registry.Get(q.Target.Host) |
| 318 | if err != nil || before == nil { |
| 319 | t.Fatalf("row before recovery: %v %v", before, err) |
| 320 | } |
| 321 | if err := root.recoverRootInventory(t.Context()); err != nil { |
| 322 | t.Fatal(err) |
| 323 | } |
| 324 | after, err := root.Reconcile.Registry.Get(q.Target.Host) |
| 325 | if err != nil || after == nil || before.ClaimOwnership != after.ClaimOwnership { |
| 326 | t.Fatalf("recovery reminted row: %v %v", after, err) |
| 327 | } |
| 328 | view, err := root.Stop.Coordinator.Guard.Root.View(t.Context(), q.Target.Host, hostguard.Captured{Incarnation: q.Target.Incarnation, Stamp: q.Target.Stamp}) |
| 329 | if err != nil || view.Installation.InventoryRevision != 1 || view.Installation.Registration.RowID != before.ClaimRowID { |
| 330 | t.Fatalf("root recovery: %+v %v", view, err) |
| 331 | } |
| 332 | } |
| 333 | |
| 334 | func TestPlan14FirstRegistrationUsesExistingInstallAuthority(t *testing.T) { |
| 335 | t.Parallel() |
| 336 | for _, mode := range []string{"never started", "writer", "missing writer", "foreign writer", "replacement", "existing install"} { |
| 337 | t.Run(mode, func(t *testing.T) { |
| 338 | kind, phase := hostclaim.KindServing, "" |
| 339 | if mode == "replacement" || mode == "existing install" { |
| 340 | kind, phase = hostclaim.KindRelocating, hostclaim.PhaseInstalled |
| 341 | } |
| 342 | f, root, q := plan14InstallRegistration(t, kind, phase) |
| 343 | o, err := hostclaim.Update(t.Context(), f.claims, f.read(), func(c *hostclaim.Claim) error { |
| 344 | switch mode { |
| 345 | case "writer": |
| 346 | c.Holder.WriterBox, c.Holder.Session = "box-1", 1 |
| 347 | case "missing writer": |
| 348 | c.Holder.WriterBox, c.Holder.Session = "", 1 |
| 349 | case "foreign writer": |
| 350 | c.Holder.WriterBox, c.Holder.Session = "box-0", 1 |
| 351 | case "existing install": |
| 352 | c.Op.Journal["local_install"] = "already-installed" |
| 353 | } |
| 354 | return nil |
| 355 | }) |
| 356 | if err != nil { |
| 357 | t.Fatal(err) |
| 358 | } |
| 359 | q.Target.ClaimETag, q.Target.Session = o.ETag, o.Claim.Holder.Session |
| 360 | q = plan14PreparedInventory(t, root, q) |
| 361 | _, err = root.commitRootInventory(t.Context(), q) |
| 362 | refuse := mode == "missing writer" || mode == "foreign writer" || mode == "existing install" |
| 363 | if (err != nil) != refuse { |
| 364 | t.Fatalf("refuse=%v: %v", refuse, err) |
| 365 | } |
| 366 | }) |
| 367 | } |
| 368 | } |
| 369 | |