| 1 | package park |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "errors" |
| 6 | "fmt" |
| 7 | "net/url" |
| 8 | "reflect" |
| 9 | "strings" |
| 10 | "testing" |
| 11 | "time" |
| 12 | |
| 13 | "github.com/carlosframework/platform/internal/hostclaim" |
| 14 | "github.com/carlosframework/platform/internal/lease" |
| 15 | "github.com/carlosframework/platform/internal/lease/leasetest" |
| 16 | "github.com/carlosframework/platform/internal/release" |
| 17 | ) |
| 18 | |
| 19 | type testLease struct { |
| 20 | released int |
| 21 | failed bool |
| 22 | } |
| 23 | |
| 24 | func (l *testLease) Check(context.Context) error { |
| 25 | if l.failed { |
| 26 | return ErrChanged |
| 27 | } |
| 28 | return nil |
| 29 | } |
| 30 | func (l *testLease) Release(context.Context) error { l.released++; return nil } |
| 31 | func (*testLease) Close() {} |
| 32 | |
| 33 | type testLeases struct{ held *testLease } |
| 34 | |
| 35 | func (l testLeases) Acquire(context.Context, string, string) (HeldLease, error) { return l.held, nil } |
| 36 | |
| 37 | type fixture struct { |
| 38 | e Engine |
| 39 | b *leasetest.VersionedBucket |
| 40 | w Work |
| 41 | now time.Time |
| 42 | ob hostclaim.Obligation |
| 43 | held *testLease |
| 44 | } |
| 45 | |
| 46 | func newFixture(t *testing.T) *fixture { |
| 47 | t.Helper() |
| 48 | ctx := t.Context() |
| 49 | now := time.Date(2026, 10, 5, 12, 0, 0, 0, time.UTC) |
| 50 | host := "shop.bab.example.com" |
| 51 | inc, err := hostclaim.Incarnation("bab", "jam", host, now) |
| 52 | if err != nil { |
| 53 | t.Fatal(err) |
| 54 | } |
| 55 | b := leasetest.NewVersionedBucket() |
| 56 | b.AddBucket("data") |
| 57 | b.PageSize = 1 |
| 58 | f := &fixture{b: b, w: Work{host, "ob", "op"}, now: now.Add(16 * time.Minute), held: &testLease{}} |
| 59 | f.ob = hostclaim.Obligation{ID: "ob", Op: "op", Kind: "park-replica", Region: "ie", Retirement: "rt1", Owner: "region:ie", Park: &hostclaim.ParkScope{Location: "instances/" + host + "/" + inc, Incarnation: inc, EvidenceAt: now, RequestLimit: time.Second}} |
| 60 | st := release.NewDirStore(t.TempDir()) |
| 61 | _, err = hostclaim.Update(ctx, st, hostclaim.Observed{Host: host, Absent: true}, func(c *hostclaim.Claim) error { |
| 62 | c.Holder = &hostclaim.Holder{Family: "instance", Key: "apps/bab/jam/instances/" + host + ".json", Account: "bab", App: "jam", Incarnation: inc, CreatedAt: now, Location: f.ob.Park.Location, Stamp: 1} |
| 63 | c.Op = hostclaim.Op{Kind: hostclaim.KindParking, ID: "op"} |
| 64 | c.Obligations = []hostclaim.Obligation{f.ob} |
| 65 | c.Touched = hostclaim.Touched{Complete: true, Regions: []string{"ie"}} |
| 66 | return nil |
| 67 | }) |
| 68 | if err != nil { |
| 69 | t.Fatal(err) |
| 70 | } |
| 71 | serial := 1 |
| 72 | f.e = Engine{Store: b, Claims: st, Bucket: "data", Region: "ie", Leases: testLeases{f.held}, Now: func() time.Time { return f.now }, NewID: func() string { serial++; return fmt.Sprintf("rt%d", serial) }, Authorize: func(context.Context, hostclaim.Observed, hostclaim.Obligation) error { return nil }, View: func(_ context.Context, o hostclaim.Observed) (View, error) { |
| 73 | return View{ClaimETag: o.ETag, Region: "ie", AuthoritativeRegion: "uk"}, nil |
| 74 | }} |
| 75 | return f |
| 76 | } |
| 77 | func (f *fixture) seed() []lease.ObjectVersion { |
| 78 | p := f.ob.Park.Location + "/" |
| 79 | f.b.Put("data", p+"0000/a", []byte("old")) |
| 80 | f.b.Put("data", p+"0000/a", []byte("new")) |
| 81 | f.b.BareDelete("data", p+"0000/a") |
| 82 | f.b.Put("data", p+"lock.json", []byte("lease")) |
| 83 | vs, _ := f.e.list(context.Background(), f.w.Host, f.ob.Park.Location) |
| 84 | return vs |
| 85 | } |
| 86 | func (f *fixture) done(t *testing.T) { |
| 87 | t.Helper() |
| 88 | obs, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 89 | if err != nil { |
| 90 | t.Fatal(err) |
| 91 | } |
| 92 | if len(obs.Claim.Obligations) != 0 { |
| 93 | t.Fatal("obligation still open") |
| 94 | } |
| 95 | if got := obs.Claim.QuietAfter["ie"]; !got.Equal(f.ob.Park.EvidenceAt.Add(SignatureAge + time.Second)) { |
| 96 | t.Fatal(got) |
| 97 | } |
| 98 | vs, err := f.e.list(t.Context(), f.w.Host, f.ob.Park.Location) |
| 99 | if err != nil || len(vs) != 0 { |
| 100 | t.Fatalf("residue: %v %v", vs, err) |
| 101 | } |
| 102 | } |
| 103 | func TestRemoteCrashResume(t *testing.T) { |
| 104 | for _, tc := range []struct { |
| 105 | step string |
| 106 | n int |
| 107 | }{{"copy", 1}, {"copy", 2}, {"manifest", 1}, {"delete", 1}, {"delete", 2}, {"delete", 3}, {"verify", 1}, {"before-completion", 1}, {"completion", 1}} { |
| 108 | t.Run(fmt.Sprintf("%s-%d", tc.step, tc.n), func(t *testing.T) { |
| 109 | f := newFixture(t) |
| 110 | source := f.seed() |
| 111 | stop := errors.New("crash") |
| 112 | n := 0 |
| 113 | f.e.After = func(s string) error { |
| 114 | if s == tc.step { |
| 115 | n++ |
| 116 | if n == tc.n { |
| 117 | return stop |
| 118 | } |
| 119 | } |
| 120 | return nil |
| 121 | } |
| 122 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, stop) { |
| 123 | t.Fatalf("crash: %v", err) |
| 124 | } |
| 125 | if tc.step != "completion" && f.held.released != 0 { |
| 126 | t.Fatal("released before ack") |
| 127 | } |
| 128 | f.e.After = nil |
| 129 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 130 | t.Fatal(err) |
| 131 | } |
| 132 | f.done(t) |
| 133 | m, err := f.e.manifest(t.Context(), f.w, f.ob) |
| 134 | if err != nil || m == nil || len(m.Delete) != len(source) || len(m.Archive) != 2 { |
| 135 | t.Fatalf("manifest: %#v %v", m, err) |
| 136 | } |
| 137 | for _, a := range m.Archive { |
| 138 | body, err := f.b.GetVersion(t.Context(), "data", a.Destination.Key, a.Destination.VersionID) |
| 139 | if err != nil || int64(len(body)) != a.Source.Size { |
| 140 | t.Fatalf("copy: %q %v", body, err) |
| 141 | } |
| 142 | } |
| 143 | if f.b.LatestID("data", f.ob.Park.Location+"/lock.json") == "" { |
| 144 | t.Fatal("lock deleted") |
| 145 | } |
| 146 | }) |
| 147 | } |
| 148 | } |
| 149 | func TestStorageFailuresKeepObligation(t *testing.T) { |
| 150 | for _, op := range []string{"list", "get", "copy", "put", "delete"} { |
| 151 | t.Run(op, func(t *testing.T) { |
| 152 | f := newFixture(t) |
| 153 | f.seed() |
| 154 | if op == "get" { |
| 155 | f.e.After = func(s string) error { |
| 156 | if s == "manifest" { |
| 157 | return ErrPending |
| 158 | } |
| 159 | return nil |
| 160 | } |
| 161 | _ = f.e.Run(t.Context(), f.w) |
| 162 | f.e.After = nil |
| 163 | } |
| 164 | failure := errors.New("injected") |
| 165 | f.b.FailOn = func(got, key string) error { |
| 166 | if got == op { |
| 167 | return failure |
| 168 | } |
| 169 | return nil |
| 170 | } |
| 171 | if err := f.e.Run(t.Context(), f.w); err == nil { |
| 172 | t.Fatal("failure acknowledged") |
| 173 | } |
| 174 | obs, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 175 | if err != nil || len(obs.Claim.Obligations) != 1 || len(obs.Claim.QuietAfter) != 0 { |
| 176 | t.Fatalf("claim: %v %v", obs, err) |
| 177 | } |
| 178 | f.b.FailOn = nil |
| 179 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 180 | t.Fatal(err) |
| 181 | } |
| 182 | f.done(t) |
| 183 | }) |
| 184 | } |
| 185 | } |
| 186 | func TestPageFailureDoesNotAct(t *testing.T) { |
| 187 | f := newFixture(t) |
| 188 | f.seed() |
| 189 | before := f.b.Snapshot() |
| 190 | f.b.FailListAfterPages = 1 |
| 191 | if err := f.e.Run(t.Context(), f.w); err == nil { |
| 192 | t.Fatal("short listing accepted") |
| 193 | } |
| 194 | if f.b.Snapshot() != before { |
| 195 | t.Fatal("mutated on listing failure") |
| 196 | } |
| 197 | } |
| 198 | func TestHorizonLatePut(t *testing.T) { |
| 199 | f := newFixture(t) |
| 200 | f.seed() |
| 201 | deadline := f.ob.Park.EvidenceAt.Add(SignatureAge + time.Second) |
| 202 | f.now = deadline |
| 203 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, ErrPending) { |
| 204 | t.Fatal(err) |
| 205 | } |
| 206 | late := f.b.Put("data", f.ob.Park.Location+"/0000/late", []byte("late")) |
| 207 | f.now = deadline.Add(time.Nanosecond) |
| 208 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, ErrPending) { |
| 209 | t.Fatal(err) |
| 210 | } |
| 211 | obs, _ := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 212 | if obs.Claim.Obligations[0].Retirement == f.ob.Retirement { |
| 213 | t.Fatal("manifest reused") |
| 214 | } |
| 215 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 216 | t.Fatal(err) |
| 217 | } |
| 218 | f.done(t) |
| 219 | if _, err := f.b.GetVersion(t.Context(), "data", f.ob.Park.Location+"/0000/late", late); !errors.Is(err, lease.ErrMirrorNotFound) { |
| 220 | t.Fatal(err) |
| 221 | } |
| 222 | } |
| 223 | func TestVerificationListingStartsAfterHorizon(t *testing.T) { |
| 224 | f := newFixture(t) |
| 225 | f.seed() |
| 226 | deadline := f.ob.Park.EvidenceAt.Add(SignatureAge + time.Second) |
| 227 | f.now = deadline |
| 228 | f.e.After = func(s string) error { |
| 229 | if s == "verify" { |
| 230 | f.now = deadline.Add(time.Hour) |
| 231 | } |
| 232 | return nil |
| 233 | } |
| 234 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, ErrPending) { |
| 235 | t.Fatal(err) |
| 236 | } |
| 237 | } |
| 238 | func TestManifestRetryDoesNotCopy(t *testing.T) { |
| 239 | f := newFixture(t) |
| 240 | f.seed() |
| 241 | f.e.After = func(s string) error { |
| 242 | if s == "delete" { |
| 243 | return ErrPending |
| 244 | } |
| 245 | return nil |
| 246 | } |
| 247 | _ = f.e.Run(t.Context(), f.w) |
| 248 | f.e.After = nil |
| 249 | f.b.FailOn = func(op, key string) error { |
| 250 | if op == "copy" { |
| 251 | t.Error("retry copied") |
| 252 | return ErrUnsafe |
| 253 | } |
| 254 | return nil |
| 255 | } |
| 256 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 257 | t.Fatal(err) |
| 258 | } |
| 259 | } |
| 260 | func TestCASConflictDoesNotAcknowledge(t *testing.T) { |
| 261 | f := newFixture(t) |
| 262 | f.seed() |
| 263 | f.e.After = func(s string) error { |
| 264 | if s == "before-completion" { |
| 265 | obs, _ := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 266 | _, err := hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { c.Lost = append(c.Lost, "other-box"); return nil }) |
| 267 | return err |
| 268 | } |
| 269 | return nil |
| 270 | } |
| 271 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, hostclaim.ErrConflict) { |
| 272 | t.Fatal(err) |
| 273 | } |
| 274 | if f.held.released != 0 { |
| 275 | t.Fatal("released unacknowledged lease") |
| 276 | } |
| 277 | f.e.After = nil |
| 278 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 279 | t.Fatal(err) |
| 280 | } |
| 281 | } |
| 282 | func TestLiveGenerationCannotBeParked(t *testing.T) { |
| 283 | f := newFixture(t) |
| 284 | obs, _ := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 285 | _, err := hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { |
| 286 | c.Op = hostclaim.Op{Kind: hostclaim.KindServing, Exec: "box"} |
| 287 | return nil |
| 288 | }) |
| 289 | if err != nil { |
| 290 | t.Fatal(err) |
| 291 | } |
| 292 | f.seed() |
| 293 | f.e.View = func(_ context.Context, o hostclaim.Observed) (View, error) { |
| 294 | return View{ClaimETag: o.ETag, Region: "ie", AuthoritativeRegion: "ie", MarkerKnown: true, MarkerGeneration: hostclaim.RootGeneration, GenerationOf: func(Ref) (string, bool) { return hostclaim.RootGeneration, true }}, nil |
| 295 | } |
| 296 | before := f.b.Snapshot() |
| 297 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, ErrUnsafe) { |
| 298 | t.Fatal(err) |
| 299 | } |
| 300 | if before != f.b.Snapshot() { |
| 301 | t.Fatal("live data changed") |
| 302 | } |
| 303 | } |
| 304 | |
| 305 | func TestRemoteMarkerDisagreementIsLocationScoped(t *testing.T) { |
| 306 | for _, scenario := range []string{"old-incarnation", "live-location", "ancestor-scope", "ancestor-holder", "unknown-region", "unknown-marker", "unknown-generations", "unknown-pinned-generation", "view-error", "authorization-error"} { |
| 307 | t.Run(scenario, func(t *testing.T) { |
| 308 | f, obs, view, _ := sweepFixture(t) |
| 309 | view.MarkerGeneration = "unrelated" |
| 310 | f.ob.Op = "session:3" |
| 311 | f.w.Op = f.ob.Op |
| 312 | if scenario != "live-location" { |
| 313 | f.ob.Park.Incarnation = strings.Repeat("a", 64) |
| 314 | f.ob.Park.Location = "instances/" + f.w.Host + "/" + f.ob.Park.Incarnation |
| 315 | } |
| 316 | if scenario == "ancestor-scope" { |
| 317 | f.ob.Park.Location = "instances/" + f.w.Host |
| 318 | } |
| 319 | _, err := hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { |
| 320 | c.Obligations = []hostclaim.Obligation{f.ob} |
| 321 | if scenario == "ancestor-holder" { |
| 322 | c.Holder.Location = "instances/" + f.w.Host |
| 323 | } |
| 324 | return nil |
| 325 | }) |
| 326 | if err != nil { |
| 327 | t.Fatal(err) |
| 328 | } |
| 329 | source := f.seed() |
| 330 | before := f.b.Snapshot() |
| 331 | liveBefore, err := f.e.list(t.Context(), f.w.Host, obs.Claim.Holder.Location) |
| 332 | if err != nil { |
| 333 | t.Fatal(err) |
| 334 | } |
| 335 | switch scenario { |
| 336 | case "unknown-region": |
| 337 | view.AuthoritativeRegion = "" |
| 338 | case "unknown-marker": |
| 339 | view.MarkerKnown = false |
| 340 | case "unknown-generations": |
| 341 | view.GenerationOf = nil |
| 342 | case "unknown-pinned-generation": |
| 343 | view.GenerationOf = func(Ref) (string, bool) { return "", false } |
| 344 | } |
| 345 | f.e.View = func(_ context.Context, o hostclaim.Observed) (View, error) { |
| 346 | v := view |
| 347 | v.ClaimETag = o.ETag |
| 348 | if scenario == "view-error" { |
| 349 | return v, ErrUnsafe |
| 350 | } |
| 351 | return v, nil |
| 352 | } |
| 353 | authorized := false |
| 354 | f.e.Authorize = func(_ context.Context, _ hostclaim.Observed, o hostclaim.Obligation) error { |
| 355 | authorized = true |
| 356 | if !reflect.DeepEqual(o, f.ob) || scenario == "authorization-error" { |
| 357 | return ErrUnsafe |
| 358 | } |
| 359 | return nil |
| 360 | } |
| 361 | err = f.e.Run(t.Context(), f.w) |
| 362 | if !authorized { |
| 363 | t.Fatal("scope was not authorized") |
| 364 | } |
| 365 | if scenario != "old-incarnation" { |
| 366 | if !errors.Is(err, ErrUnsafe) { |
| 367 | t.Fatalf("Run = %v, want refusal", err) |
| 368 | } |
| 369 | if f.b.Snapshot() != before { |
| 370 | t.Fatal("refused work changed storage") |
| 371 | } |
| 372 | pending, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 373 | if err != nil { |
| 374 | t.Fatal(err) |
| 375 | } |
| 376 | if len(pending.Claim.Obligations) != 1 || len(pending.Claim.QuietAfter) != 0 || f.held.released != 0 { |
| 377 | t.Fatal("refused work acknowledged") |
| 378 | } |
| 379 | return |
| 380 | } |
| 381 | if err != nil { |
| 382 | t.Fatalf("old incarnation refused by live marker disagreement: %v", err) |
| 383 | } |
| 384 | f.done(t) |
| 385 | m, err := f.e.manifest(t.Context(), f.w, f.ob) |
| 386 | if err != nil || m == nil || !reflect.DeepEqual(m.Delete, source) || len(m.Archive) != 2 { |
| 387 | t.Fatalf("old incarnation manifest: %+v %v", m, err) |
| 388 | } |
| 389 | for _, copy := range m.Archive { |
| 390 | body, err := f.b.GetVersion(t.Context(), "data", copy.Destination.Key, copy.Destination.VersionID) |
| 391 | if err != nil || (string(body) != "old" && string(body) != "new") { |
| 392 | t.Fatalf("archive: %q %v", body, err) |
| 393 | } |
| 394 | } |
| 395 | liveAfter, err := f.e.list(t.Context(), f.w.Host, obs.Claim.Holder.Location) |
| 396 | if err != nil || !reflect.DeepEqual(liveBefore, liveAfter) { |
| 397 | t.Fatalf("live location changed: %v", err) |
| 398 | } |
| 399 | if f.held.released != 1 { |
| 400 | t.Fatal("completed work did not release its lease") |
| 401 | } |
| 402 | }) |
| 403 | } |
| 404 | } |
| 405 | |
| 406 | func TestBoundScopeAndQuietAfter(t *testing.T) { |
| 407 | f := newFixture(t) |
| 408 | obs, _ := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 409 | later := f.now.Add(time.Hour) |
| 410 | _, err := hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { |
| 411 | c.Obligations[0].Park = nil |
| 412 | c.QuietAfter = map[string]time.Time{"ie": later} |
| 413 | return nil |
| 414 | }) |
| 415 | if err != nil { |
| 416 | t.Fatal(err) |
| 417 | } |
| 418 | if err := f.e.Bind(t.Context(), f.w, *f.ob.Park); err != nil { |
| 419 | t.Fatal(err) |
| 420 | } |
| 421 | bad := *f.ob.Park |
| 422 | bad.EvidenceAt = bad.EvidenceAt.Add(time.Second) |
| 423 | if err := f.e.Bind(t.Context(), f.w, bad); !errors.Is(err, ErrChanged) { |
| 424 | t.Fatal(err) |
| 425 | } |
| 426 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 427 | t.Fatal(err) |
| 428 | } |
| 429 | obs, _ = hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 430 | if !obs.Claim.QuietAfter["ie"].Equal(later) { |
| 431 | t.Fatal("quiet_after regressed") |
| 432 | } |
| 433 | } |
| 434 | func TestBindSurvivesSerialization(t *testing.T) { |
| 435 | f := newFixture(t) |
| 436 | obs, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 437 | if err != nil { |
| 438 | t.Fatal(err) |
| 439 | } |
| 440 | _, err = hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { |
| 441 | c.Obligations[0].Park = nil |
| 442 | return nil |
| 443 | }) |
| 444 | if err != nil { |
| 445 | t.Fatal(err) |
| 446 | } |
| 447 | scope := *f.ob.Park |
| 448 | scope.EvidenceAt = time.Now() // Carries a monotonic reading that JSON drops. |
| 449 | if err := f.e.Bind(t.Context(), f.w, scope); err != nil { |
| 450 | t.Fatal(err) |
| 451 | } |
| 452 | before, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 453 | if err != nil { |
| 454 | t.Fatal(err) |
| 455 | } |
| 456 | for _, at := range []time.Time{scope.EvidenceAt, scope.EvidenceAt.UTC()} { |
| 457 | same := scope |
| 458 | same.EvidenceAt = at |
| 459 | if err := f.e.Bind(t.Context(), f.w, same); err != nil { |
| 460 | t.Fatal(err) |
| 461 | } |
| 462 | } |
| 463 | for _, change := range []func(*hostclaim.ParkScope){ |
| 464 | func(p *hostclaim.ParkScope) { p.Location += "/other" }, |
| 465 | func(p *hostclaim.ParkScope) { p.Incarnation = strings.Repeat("f", 64) }, |
| 466 | func(p *hostclaim.ParkScope) { p.RequestLimit += time.Second }, |
| 467 | func(p *hostclaim.ParkScope) { p.EvidenceAt = p.EvidenceAt.Add(time.Nanosecond) }, |
| 468 | } { |
| 469 | bad := scope |
| 470 | change(&bad) |
| 471 | if err := f.e.Bind(t.Context(), f.w, bad); !errors.Is(err, ErrChanged) { |
| 472 | t.Fatal(err) |
| 473 | } |
| 474 | } |
| 475 | after, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 476 | if err != nil || after.ETag != before.ETag { |
| 477 | t.Fatalf("bind changed persisted scope: %v", err) |
| 478 | } |
| 479 | } |
| 480 | |
| 481 | func TestLegacyLocationDoesNotIncludeIncarnationChildren(t *testing.T) { |
| 482 | f := newFixture(t) |
| 483 | key := f.ob.Park.Location + "/0000/data" |
| 484 | f.b.Put("data", key, []byte("new incarnation")) |
| 485 | vs, err := f.e.list(t.Context(), f.w.Host, "instances/"+f.w.Host) |
| 486 | if err != nil || len(vs) != 0 { |
| 487 | t.Fatalf("%v %v", vs, err) |
| 488 | } |
| 489 | for _, v := range []string{"a/b", "a@b", "a+b", "a b", "a?b"} { |
| 490 | key := destination(f.w.Host, f.ob, lease.ObjectVersion{Key: key, VersionID: v}) |
| 491 | if strings.HasSuffix(key, "@"+v) { |
| 492 | t.Errorf("version not escaped: %s", v) |
| 493 | } |
| 494 | } |
| 495 | } |
| 496 | |
| 497 | func TestDeleteParksStoppedAuthoritativeLocation(t *testing.T) { |
| 498 | f := newFixture(t) |
| 499 | f.seed() |
| 500 | f.e.View = func(_ context.Context, o hostclaim.Observed) (View, error) { |
| 501 | return View{ClaimETag: o.ETag, Region: "ie", AuthoritativeRegion: "ie"}, nil |
| 502 | } |
| 503 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 504 | t.Fatal(err) |
| 505 | } |
| 506 | f.done(t) |
| 507 | } |
| 508 | func TestForeignRootGenerationRetirement(t *testing.T) { |
| 509 | f := newFixture(t) |
| 510 | f.seed() |
| 511 | obs, _ := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 512 | _, err := hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { |
| 513 | c.Obligations[0].Kind = "park-generation" |
| 514 | c.Obligations[0].Generation = hostclaim.RootGeneration |
| 515 | return nil |
| 516 | }) |
| 517 | if err != nil { |
| 518 | t.Fatal(err) |
| 519 | } |
| 520 | f.e.View = func(_ context.Context, o hostclaim.Observed) (View, error) { |
| 521 | return View{ClaimETag: o.ETag, Region: "ie", AuthoritativeRegion: "uk", GenerationOf: func(Ref) (string, bool) { return "", true }}, nil |
| 522 | } |
| 523 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 524 | t.Fatal(err) |
| 525 | } |
| 526 | f.done(t) |
| 527 | } |
| 528 | |
| 529 | func TestLeaseLossAndAuthorizationRefuse(t *testing.T) { |
| 530 | for _, phase := range []string{"before-first-effect", "before-completion", "authorize"} { |
| 531 | t.Run(phase, func(t *testing.T) { |
| 532 | f := newFixture(t) |
| 533 | f.seed() |
| 534 | before := f.b.Snapshot() |
| 535 | switch phase { |
| 536 | case "before-first-effect": |
| 537 | f.held.failed = true |
| 538 | case "before-completion": |
| 539 | f.e.After = func(s string) error { |
| 540 | if s == phase { |
| 541 | f.held.failed = true |
| 542 | } |
| 543 | return nil |
| 544 | } |
| 545 | case "authorize": |
| 546 | f.e.Authorize = func(context.Context, hostclaim.Observed, hostclaim.Obligation) error { return ErrChanged } |
| 547 | } |
| 548 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, ErrChanged) { |
| 549 | t.Fatal(err) |
| 550 | } |
| 551 | obs, _ := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 552 | if len(obs.Claim.Obligations) != 1 || f.held.released != 0 { |
| 553 | t.Fatal("acknowledged without authority") |
| 554 | } |
| 555 | if phase != "before-completion" && f.b.Snapshot() != before { |
| 556 | t.Fatal("acted without authority") |
| 557 | } |
| 558 | }) |
| 559 | } |
| 560 | } |
| 561 | func TestMalformedManifestNeverDeletes(t *testing.T) { |
| 562 | f := newFixture(t) |
| 563 | f.seed() |
| 564 | _, err := f.b.PutIfAbsent(t.Context(), "data", archiveBase(f.w.Host, f.ob)+"manifest.json", []byte(`{"V":1,"Host":"other.example"}`)) |
| 565 | if err != nil { |
| 566 | t.Fatal(err) |
| 567 | } |
| 568 | before := f.b.Snapshot() |
| 569 | if err := f.e.Run(t.Context(), f.w); err == nil { |
| 570 | t.Fatal("accepted malformed manifest") |
| 571 | } |
| 572 | if before != f.b.Snapshot() { |
| 573 | t.Fatal("deleted from bad manifest") |
| 574 | } |
| 575 | } |
| 576 | |
| 577 | // Parked is what the landing's parking phase waits for: the manifest is |
| 578 | // there and every version it lists is gone. It is not the acknowledgement, |
| 579 | // which still waits for the horizon in Run. |
| 580 | func TestParkedReportsManifestAndDeletions(t *testing.T) { |
| 581 | f := newFixture(t) |
| 582 | f.seed() |
| 583 | parked := func() bool { |
| 584 | t.Helper() |
| 585 | ok, err := f.e.Parked(t.Context(), f.w) |
| 586 | if err != nil { |
| 587 | t.Fatalf("Parked: %v", err) |
| 588 | } |
| 589 | return ok |
| 590 | } |
| 591 | if parked() { |
| 592 | t.Fatal("parked before any work") |
| 593 | } |
| 594 | stop := errors.New("crash") |
| 595 | crashAt := func(step string, n int) { |
| 596 | t.Helper() |
| 597 | seen := 0 |
| 598 | f.e.After = func(s string) error { |
| 599 | if s == step { |
| 600 | seen++ |
| 601 | if seen == n { |
| 602 | return stop |
| 603 | } |
| 604 | } |
| 605 | return nil |
| 606 | } |
| 607 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, stop) { |
| 608 | t.Fatalf("crash at %s: %v", step, err) |
| 609 | } |
| 610 | f.e.After = nil |
| 611 | } |
| 612 | crashAt("manifest", 1) |
| 613 | if parked() { |
| 614 | t.Fatal("parked with a manifest and nothing deleted") |
| 615 | } |
| 616 | crashAt("delete", 1) |
| 617 | if parked() { |
| 618 | t.Fatal("parked with versions left to delete") |
| 619 | } |
| 620 | // Resume through every delete, stopping at the verification: the |
| 621 | // manifest is there and everything in it is gone, but the obligation |
| 622 | // is not acknowledged. |
| 623 | crashAt("verify", 1) |
| 624 | if !parked() { |
| 625 | t.Fatal("not parked after every listed version is gone") |
| 626 | } |
| 627 | obs, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 628 | if err != nil || len(obs.Claim.Obligations) != 1 { |
| 629 | t.Fatalf("Parked must not acknowledge: %v %v", obs.Claim.Obligations, err) |
| 630 | } |
| 631 | // A version the manifest does not list is not in the deletion set: the |
| 632 | // retirement's own verification pass handles it, Parked only reports on |
| 633 | // what the manifest promised. |
| 634 | f.b.Put("data", f.ob.Park.Location+"/0000/late", []byte("late")) |
| 635 | if !parked() { |
| 636 | t.Fatal("an unlisted late version changed the manifest's answer") |
| 637 | } |
| 638 | } |
| 639 | |
| 640 | // A key with many versions is listed once, however many of its versions are |
| 641 | // in the deletion set. |
| 642 | func TestParkedListsEachKeyOnce(t *testing.T) { |
| 643 | f := newFixture(t) |
| 644 | f.seed() |
| 645 | stop := errors.New("crash") |
| 646 | f.e.After = func(s string) error { |
| 647 | if s == "verify" { |
| 648 | return stop |
| 649 | } |
| 650 | return nil |
| 651 | } |
| 652 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, stop) { |
| 653 | t.Fatal(err) |
| 654 | } |
| 655 | f.e.After = nil |
| 656 | m, err := f.e.manifest(t.Context(), f.w, f.ob) |
| 657 | if err != nil || m == nil { |
| 658 | t.Fatal(m, err) |
| 659 | } |
| 660 | perKey := map[string]int{} |
| 661 | for _, d := range m.Delete { |
| 662 | perKey[d.Key]++ |
| 663 | } |
| 664 | if len(m.Delete) < 3 || len(perKey) != 1 { |
| 665 | t.Fatalf("the fixture must put several versions under one key: %v", perKey) |
| 666 | } |
| 667 | // The manifest and chore reads list too; count only the deletion keys. |
| 668 | lists := 0 |
| 669 | f.b.FailOn = func(op, key string) error { |
| 670 | if op == "list" && perKey[key] > 0 { |
| 671 | lists++ |
| 672 | } |
| 673 | return nil |
| 674 | } |
| 675 | if ok, err := f.e.Parked(t.Context(), f.w); !ok || err != nil { |
| 676 | t.Fatalf("parked: %v %v", ok, err) |
| 677 | } |
| 678 | f.b.FailOn = nil |
| 679 | if lists != 1 { |
| 680 | t.Fatalf("a key with %d deleted versions was listed %d times", len(m.Delete), lists) |
| 681 | } |
| 682 | } |
| 683 | |
| 684 | func TestParkedNeverGuesses(t *testing.T) { |
| 685 | f := newFixture(t) |
| 686 | f.seed() |
| 687 | stop := errors.New("crash") |
| 688 | f.e.After = func(s string) error { |
| 689 | if s == "verify" { |
| 690 | return stop |
| 691 | } |
| 692 | return nil |
| 693 | } |
| 694 | if err := f.e.Run(t.Context(), f.w); !errors.Is(err, stop) { |
| 695 | t.Fatal(err) |
| 696 | } |
| 697 | f.e.After = nil |
| 698 | boom := errors.New("offline") |
| 699 | for _, fail := range []string{"list", "get"} { |
| 700 | f.b.FailOn = func(op, key string) error { |
| 701 | if op == fail { |
| 702 | return boom |
| 703 | } |
| 704 | return nil |
| 705 | } |
| 706 | ok, err := f.e.Parked(t.Context(), f.w) |
| 707 | if ok || err == nil { |
| 708 | t.Fatalf("%s failure answered %v, %v", fail, ok, err) |
| 709 | } |
| 710 | } |
| 711 | f.b.FailOn = nil |
| 712 | if ok, err := f.e.Parked(t.Context(), f.w); !ok || err != nil { |
| 713 | t.Fatalf("recovered read: %v %v", ok, err) |
| 714 | } |
| 715 | // A chore that is not in the claim cannot be verified from its manifest. |
| 716 | if ok, err := f.e.Parked(t.Context(), Work{f.w.Host, "nope", f.w.Op}); ok || !errors.Is(err, ErrChanged) { |
| 717 | t.Fatalf("unknown chore: %v %v", ok, err) |
| 718 | } |
| 719 | // A rotated retirement has no manifest yet: not parked, and not an error. |
| 720 | obs, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 721 | if err != nil { |
| 722 | t.Fatal(err) |
| 723 | } |
| 724 | _, err = hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { |
| 725 | c.Obligations[0].Retirement = "rt-other" |
| 726 | return nil |
| 727 | }) |
| 728 | if err != nil { |
| 729 | t.Fatal(err) |
| 730 | } |
| 731 | if ok, err := f.e.Parked(t.Context(), f.w); ok || err != nil { |
| 732 | t.Fatalf("a retirement with no manifest answered %v, %v", ok, err) |
| 733 | } |
| 734 | if ok, err := (Engine{}).Parked(t.Context(), f.w); ok || !errors.Is(err, ErrUnsafe) { |
| 735 | t.Fatalf("unconfigured engine: %v %v", ok, err) |
| 736 | } |
| 737 | } |
| 738 | |
| 739 | // A move's retirement goes through the same engine as any other: queued by |
| 740 | // hostclaim.QueueLocationPark on the existing park-replica kind, archived |
| 741 | // under parked/<incarnation>/<host>/<retirement>/, and the manifest's Op is |
| 742 | // the move id (F3 reads it). |
| 743 | func TestEngineArchivesAMoveRetirementUnderParked(t *testing.T) { |
| 744 | f := newFixture(t) |
| 745 | const move = "mv_abcdefghijklmnopqrstuvwxyz" |
| 746 | scope := *f.ob.Park |
| 747 | n := 0 |
| 748 | newID := func() string { n++; return fmt.Sprintf("q%d", n) } |
| 749 | obs, err := hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 750 | if err != nil { |
| 751 | t.Fatal(err) |
| 752 | } |
| 753 | _, err = hostclaim.Update(t.Context(), f.e.Claims, obs, func(c *hostclaim.Claim) error { |
| 754 | // The source's region after the commit: the claim's landing/committed. |
| 755 | c.Op = hostclaim.Op{Kind: hostclaim.KindLanding, ID: move, Phase: "committed", Intent: map[string]string{"to": "uk"}} |
| 756 | c.Obligations = nil |
| 757 | return hostclaim.QueueLocationPark(move, "ie", "", scope, newID)(c) |
| 758 | }) |
| 759 | if err != nil { |
| 760 | t.Fatal(err) |
| 761 | } |
| 762 | obs, err = hostclaim.Read(t.Context(), f.e.Claims, f.w.Host) |
| 763 | if err != nil || len(obs.Claim.Obligations) != 1 { |
| 764 | t.Fatal(obs.Claim.Obligations, err) |
| 765 | } |
| 766 | ob := obs.Claim.Obligations[0] |
| 767 | f.ob, f.w = ob, Work{f.w.Host, ob.ID, move} |
| 768 | id := f.b.Put("data", scope.Location+"/0000/x.ltx", []byte("ltx")) |
| 769 | if err := f.e.Run(t.Context(), f.w); err != nil { |
| 770 | t.Fatal(err) |
| 771 | } |
| 772 | m, err := f.e.manifest(t.Context(), f.w, ob) |
| 773 | if err != nil || m == nil { |
| 774 | t.Fatalf("manifest: %v %v", m, err) |
| 775 | } |
| 776 | if m.Op != move || m.Retirement != ob.Retirement || m.Location != scope.Location { |
| 777 | t.Fatalf("manifest: %+v", m) |
| 778 | } |
| 779 | want := "parked/" + scope.Incarnation + "/" + f.w.Host + "/" + ob.Retirement + "/instances/0000/x.ltx@" + url.QueryEscape(id) |
| 780 | if len(m.Archive) != 1 || m.Archive[0].Destination.Key != want { |
| 781 | t.Fatalf("archive %+v, want %s", m.Archive, want) |
| 782 | } |
| 783 | f.done(t) |
| 784 | } |
| 785 | |