| 1 | package hostactors |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "encoding/json" |
| 6 | "errors" |
| 7 | "fmt" |
| 8 | "os" |
| 9 | "strings" |
| 10 | "testing" |
| 11 | "time" |
| 12 | |
| 13 | "github.com/carlosframework/platform/internal/boxid" |
| 14 | "github.com/carlosframework/platform/internal/boxid/boxidtest" |
| 15 | "github.com/carlosframework/platform/internal/release" |
| 16 | "github.com/carlosframework/platform/internal/status" |
| 17 | ) |
| 18 | |
| 19 | // enrollWorld is plan9Fixture's published deployment (boxes a, b and w, |
| 20 | // active in ie under a published host floor) with ie's group, and a box |
| 21 | // (instance i-new, reference newRef) that the group launched: its |
| 22 | // identity record is published, and nothing else. The console reads the |
| 23 | // cloud through cloud. |
| 24 | type enrollWorld struct { |
| 25 | plan9World |
| 26 | e *AutoEnroller |
| 27 | cloud *fakeMembership |
| 28 | signer *boxidtest.Signer |
| 29 | clock time.Time |
| 30 | logs []string |
| 31 | } |
| 32 | |
| 33 | func newEnrollWorld(t *testing.T, role string) *enrollWorld { |
| 34 | t.Helper() |
| 35 | w, _ := plan9Fixture(t, func(r *Roster) { |
| 36 | r.Regions = map[string]Region{"ie": {Account: "123", AWSRegion: "eu-west-1", GroupName: "routers-ie", |
| 37 | GroupARN: "arn:aws:autoscaling:eu-west-1:123:autoScalingGroup:x:autoScalingGroupName/routers-ie", Namespaces: []string{"platform"}, Pools: []string{"default"}, Role: role}} |
| 38 | }) |
| 39 | signer := boxidtest.NewSigner(t) |
| 40 | _, raw := signer.Record(t, newRef, "123", "eu-west-1", "i-new") |
| 41 | if _, err := w.d.Store.Put(context.Background(), boxid.Key(newRef), raw, ""); err != nil { |
| 42 | t.Fatal(err) |
| 43 | } |
| 44 | ew := &enrollWorld{plan9World: w, clock: time.Now().UTC(), signer: signer, |
| 45 | cloud: &fakeMembership{members: map[string]Membership{"i-new": member("routers-ie")}}} |
| 46 | ew.e = &AutoEnroller{Directory: w.d, Signer: w.sg, Evidence: ExactEvidence{PublicKey: w.d.PublicKey, Store: w.d.Store, Boxes: w.d.PrivateBoxReport}, |
| 47 | Members: ew.cloud, Certs: signer.Certs(), Namespace: "platform", Pool: "default", Now: func() time.Time { return ew.clock }, Patience: 2 * time.Minute, |
| 48 | Logf: func(format string, args ...any) { ew.logs = append(ew.logs, fmt.Sprintf(format, args...)) }} |
| 49 | return ew |
| 50 | } |
| 51 | |
| 52 | // pass runs the enrollment pass and returns new's step. |
| 53 | func (w *enrollWorld) pass(t *testing.T) string { |
| 54 | t.Helper() |
| 55 | steps, err := w.e.EnrollPass(context.Background()) |
| 56 | if err != nil { |
| 57 | t.Fatal(err) |
| 58 | } |
| 59 | for _, s := range steps { |
| 60 | if s.Box == newRef { |
| 61 | return s.Step |
| 62 | } |
| 63 | } |
| 64 | return "" |
| 65 | } |
| 66 | |
| 67 | // guardReady is the joining box, its guard armed, publishing the report |
| 68 | // every hosting box publishes. |
| 69 | func (w *enrollWorld) guardReady(t *testing.T, enrollment string, ready bool) { |
| 70 | t.Helper() |
| 71 | box := status.Box{GeneratedAt: time.Now().UTC(), Metrics: &status.BoxMetrics{Enrollment: enrollment, InstanceID: "i-new", Release: "v1.0.0", ReleaseTag: "v1.0.0", |
| 72 | Platform: "linux-amd64", BinarySHA256: w.hash, MovesProtocol: 1, BootID: "boot", FloorSHA256: w.floor.Digest, FloorGuardReady: ready, SystemChannel: "stable"}} |
| 73 | raw, _ := json.Marshal(box) |
| 74 | _, etag, _ := w.d.Store.Get(context.Background(), "status/boxes/"+newRef+".json") |
| 75 | if _, err := w.d.Store.Put(context.Background(), "status/boxes/"+newRef+".json", raw, etag); err != nil { |
| 76 | t.Fatal(err) |
| 77 | } |
| 78 | } |
| 79 | |
| 80 | func (w *enrollWorld) invitation(t *testing.T) (Invitation, bool) { |
| 81 | t.Helper() |
| 82 | ctx := context.Background() |
| 83 | wire, _, err := w.d.Store.Get(ctx, "control/host-actor-invitations/"+newRef+".json") |
| 84 | if errors.Is(err, os.ErrNotExist) { |
| 85 | return Invitation{}, false |
| 86 | } |
| 87 | if err != nil { |
| 88 | t.Fatal(err) |
| 89 | } |
| 90 | var got Invitation |
| 91 | if err := release.VerifyRootEnvelope(w.d.PublicKey, wire, &got); err != nil { |
| 92 | t.Fatal(err) |
| 93 | } |
| 94 | // Read only, never Invite (get-or-create): the invitation verifies |
| 95 | // for the box's record, and nothing is written. |
| 96 | raw, err := boxid.Load(ctx, w.d.Store, newRef) |
| 97 | if err != nil { |
| 98 | t.Fatal(err) |
| 99 | } |
| 100 | nonce, err := w.d.enrollmentNonce(ctx, EnrollmentRequest{Reference: newRef, Namespace: "platform"}, raw) |
| 101 | if err != nil || nonce != got.Enrollment { |
| 102 | t.Fatalf("invitation %+v: %v", got, err) |
| 103 | } |
| 104 | return got, true |
| 105 | } |
| 106 | |
| 107 | func (w *enrollWorld) attempt(t *testing.T) EnrollmentAttempt { |
| 108 | t.Helper() |
| 109 | o, err := w.d.Read(context.Background()) |
| 110 | if err != nil { |
| 111 | t.Fatal(err) |
| 112 | } |
| 113 | return o.Roster.EnrollmentAttempts[newRef] |
| 114 | } |
| 115 | |
| 116 | // The whole sequence (spec 2 §5 step 5, §11): attempt, the console's own |
| 117 | // membership read and the invitation, guard-ready, the console's second |
| 118 | // read and the enrollment. The deployment's floor is published, so the |
| 119 | // enrolling write passes membership.go's rule only because the guard-ready |
| 120 | // report is there; the box takes its group's role in it. |
| 121 | func TestAutomaticEnrollmentFromRecordToRoster(t *testing.T) { |
| 122 | w := newEnrollWorld(t, RoleLegacy) |
| 123 | if got := w.pass(t); got != "began an attempt" { |
| 124 | t.Fatalf("first pass: %q", got) |
| 125 | } |
| 126 | if _, ok := w.invitation(t); ok || len(w.cloud.calls) != 0 { |
| 127 | t.Fatalf("invited before membership was read: calls %v", w.cloud.calls) |
| 128 | } |
| 129 | if got := w.pass(t); got != "invited" { |
| 130 | t.Fatalf("second pass: %q", got) |
| 131 | } |
| 132 | inv, ok := w.invitation(t) |
| 133 | if !ok || len(w.cloud.calls) != 1 { |
| 134 | t.Fatalf("invitation %v, calls %v", ok, w.cloud.calls) |
| 135 | } |
| 136 | // Invited, but the guard is not armed: nothing more, and no read. |
| 137 | // First no report at all, then a report from a box whose guard is not |
| 138 | // ready. |
| 139 | if got := w.pass(t); got != "waiting for its guard-ready report" { |
| 140 | t.Fatalf("invited, no report: %q", got) |
| 141 | } |
| 142 | w.guardReady(t, inv.Enrollment, false) |
| 143 | if got := w.pass(t); got != "waiting for its guard-ready report" || len(w.cloud.calls) != 1 { |
| 144 | t.Fatalf("invited, guard not ready: %q, calls %v", got, w.cloud.calls) |
| 145 | } |
| 146 | w.guardReady(t, inv.Enrollment, true) |
| 147 | w.clock = w.clock.Add(5 * time.Minute) |
| 148 | if got := w.pass(t); !strings.HasPrefix(got, "enrolled in ie") { |
| 149 | t.Fatalf("ready: %q", got) |
| 150 | } |
| 151 | // The enrollment read membership again, itself. |
| 152 | if len(w.cloud.calls) != 2 || len(w.logs) != 2 || !strings.Contains(w.logs[1], "attempt "+w.attempt(t).AttemptID+", instance i-new, request ids asg-req, ec2-req") || !strings.HasSuffix(w.logs[1], "): enrolled") { |
| 153 | t.Fatalf("calls %v, logs %q", w.cloud.calls, w.logs) |
| 154 | } |
| 155 | o, err := w.d.Read(context.Background()) |
| 156 | if err != nil { |
| 157 | t.Fatal(err) |
| 158 | } |
| 159 | b, ok := o.Roster.Boxes[newRef] |
| 160 | if !ok || b.State != "active" || b.Enrollment != inv.Enrollment || b.Cloud.InstanceID != "i-new" || b.Group != "routers-ie" || b.Role != RoleLegacy || b.Namespace != "platform" || b.Pool != "default" { |
| 161 | t.Fatalf("roster entry %+v", b) |
| 162 | } |
| 163 | if !o.Roster.EnrollmentAttempts[newRef].Consumed { |
| 164 | t.Fatal("the attempt was not consumed") |
| 165 | } |
| 166 | // Enrolled: the pass has nothing left to do for it. |
| 167 | if got := w.pass(t); got != "" { |
| 168 | t.Fatalf("enrolled box stepped again: %q", got) |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | // R1: an instance in the deployment's account that the cloud does not |
| 173 | // place in the region's group is never invited and never enrolled, however |
| 174 | // long it waits; and one that leaves the group after its invitation is not |
| 175 | // enrolled on a guard-ready report. |
| 176 | func TestAnInstanceOutsideTheGroupIsNeverInvitedOrEnrolled(t *testing.T) { |
| 177 | for name, m := range map[string]Membership{ |
| 178 | "another group": member("routers-uk"), |
| 179 | "no group": {Owner: "123", InstanceState: "running", AvailabilityZoneRegion: "eu-west-1", RequestIDs: []string{"a", "b"}}, |
| 180 | } { |
| 181 | t.Run(name, func(t *testing.T) { |
| 182 | w := newEnrollWorld(t, "") |
| 183 | w.cloud.members["i-new"] = m |
| 184 | for i := 0; i < 12; i++ { |
| 185 | w.pass(t) |
| 186 | w.clock = w.clock.Add(time.Minute) |
| 187 | } |
| 188 | if _, ok := w.invitation(t); ok { |
| 189 | t.Fatal("an instance outside the group was invited") |
| 190 | } |
| 191 | if o, _ := w.d.Read(context.Background()); o.Roster.Boxes[newRef].Enrollment != "" { |
| 192 | t.Fatal("an instance outside the group was enrolled") |
| 193 | } |
| 194 | }) |
| 195 | } |
| 196 | w := newEnrollWorld(t, "") |
| 197 | w.pass(t) |
| 198 | if got := w.pass(t); got != "invited" { |
| 199 | t.Fatal(got) |
| 200 | } |
| 201 | inv, _ := w.invitation(t) |
| 202 | w.guardReady(t, inv.Enrollment, true) |
| 203 | w.cloud.members["i-new"] = func() Membership { m := member("routers-ie"); m.LifecycleState = "Detaching"; return m }() |
| 204 | if got := w.pass(t); !strings.Contains(got, "not confirmed") { |
| 205 | t.Fatalf("left the group: %q", got) |
| 206 | } |
| 207 | if o, _ := w.d.Read(context.Background()); o.Roster.Boxes[newRef].Enrollment != "" { |
| 208 | t.Fatal("an instance that left the group was enrolled") |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | // R1 (#7): a read that fails is unknown. The attempt stands, its wait does |
| 213 | // not grow, and the next read that answers goes on from it. |
| 214 | func TestAMembershipReadThatFailsKeepsTheAttempt(t *testing.T) { |
| 215 | w := newEnrollWorld(t, "") |
| 216 | w.pass(t) |
| 217 | first := w.attempt(t).AttemptID |
| 218 | w.cloud.err = errors.New("Throttling: rate exceeded") |
| 219 | for i := 0; i < 10; i++ { |
| 220 | w.clock = w.clock.Add(time.Hour) |
| 221 | if got := w.pass(t); !strings.Contains(got, "could not be read") { |
| 222 | t.Fatalf("failed read: %q", got) |
| 223 | } |
| 224 | } |
| 225 | if got := w.attempt(t).AttemptID; got != first || w.e.attempts[newRef].tries != 1 { |
| 226 | t.Fatalf("a failed read superseded the attempt (%s, was %s) or grew its wait (%d)", got, first, w.e.attempts[newRef].tries) |
| 227 | } |
| 228 | w.cloud.err = nil |
| 229 | if got := w.pass(t); got != "invited" || w.attempt(t).AttemptID != first { |
| 230 | t.Fatalf("after the outage: %q", got) |
| 231 | } |
| 232 | } |
| 233 | |
| 234 | // stalledMembership is a describe whose connection stalls for one |
| 235 | // instance: Member returns only when its context ends. Every other |
| 236 | // instance is answered by the cloud underneath. |
| 237 | type stalledMembership struct { |
| 238 | stalled string |
| 239 | cloud *fakeMembership |
| 240 | deadline bool |
| 241 | } |
| 242 | |
| 243 | func (s *stalledMembership) Member(ctx context.Context, awsRegion, instance string) (Membership, error) { |
| 244 | if instance != s.stalled { |
| 245 | return s.cloud.Member(ctx, awsRegion, instance) |
| 246 | } |
| 247 | _, s.deadline = ctx.Deadline() |
| 248 | <-ctx.Done() |
| 249 | return Membership{}, ctx.Err() |
| 250 | } |
| 251 | |
| 252 | // A membership read that stalls is cut off at storeCallTimeout and is |
| 253 | // unknown: the pass returns within the bound, the stalled box's attempt |
| 254 | // stands with its wait unchanged, and the box after it in the same pass is |
| 255 | // still invited. |
| 256 | func TestAStalledMembershipReadIsCutOffAndThePassGoesOn(t *testing.T) { |
| 257 | was := storeCallTimeout |
| 258 | storeCallTimeout = 100 * time.Millisecond |
| 259 | t.Cleanup(func() { storeCallTimeout = was }) |
| 260 | w := newEnrollWorld(t, "") |
| 261 | // A second box the group launched, whose reference sorts after |
| 262 | // newRef's, so it waits in the pass behind the stalled read. |
| 263 | twoRef := GroupBoxReference("123", "i-two") |
| 264 | if twoRef <= newRef { |
| 265 | t.Fatalf("%s does not sort after %s", twoRef, newRef) |
| 266 | } |
| 267 | _, raw := w.signer.Record(t, twoRef, "123", "eu-west-1", "i-two") |
| 268 | if _, err := w.d.Store.Put(context.Background(), boxid.Key(twoRef), raw, ""); err != nil { |
| 269 | t.Fatal(err) |
| 270 | } |
| 271 | w.cloud.members["i-two"] = member("routers-ie") |
| 272 | stalled := &stalledMembership{stalled: "i-new", cloud: w.cloud} |
| 273 | w.e.Members = stalled |
| 274 | steps, err := w.e.EnrollPass(context.Background()) |
| 275 | if err != nil || len(steps) != 2 || steps[0].Step != "began an attempt" || steps[1].Step != "began an attempt" { |
| 276 | t.Fatalf("first pass: %v, %v", steps, err) |
| 277 | } |
| 278 | first := w.attempt(t).AttemptID |
| 279 | |
| 280 | done := make(chan []EnrollStep, 1) |
| 281 | start := time.Now() |
| 282 | go func() { |
| 283 | steps, err := w.e.EnrollPass(context.Background()) |
| 284 | if err != nil { |
| 285 | t.Error(err) |
| 286 | } |
| 287 | done <- steps |
| 288 | }() |
| 289 | select { |
| 290 | case steps = <-done: |
| 291 | case <-time.After(10 * time.Second): |
| 292 | t.Fatal("the pass is still waiting on the stalled read") |
| 293 | } |
| 294 | if took := time.Since(start); took > storeCallTimeout+5*time.Second { |
| 295 | t.Fatalf("the pass took %s", took) |
| 296 | } |
| 297 | if !stalled.deadline { |
| 298 | t.Fatal("the read was given no deadline") |
| 299 | } |
| 300 | if len(steps) != 2 || steps[0].Box != newRef || !strings.Contains(steps[0].Step, "could not be read") || !strings.Contains(steps[0].Step, "did not finish in time") { |
| 301 | t.Fatalf("stalled box: %v", steps) |
| 302 | } |
| 303 | if got := w.attempt(t).AttemptID; got != first || w.e.attempts[newRef].tries != 1 || w.e.attempts[newRef].refused != "" { |
| 304 | t.Fatalf("the stalled read superseded the attempt (%s, was %s), grew its wait (%d) or refused it (%q)", got, first, w.e.attempts[newRef].tries, w.e.attempts[newRef].refused) |
| 305 | } |
| 306 | if _, ok := w.invitation(t); ok { |
| 307 | t.Fatal("the stalled box was invited") |
| 308 | } |
| 309 | if steps[1].Box != twoRef || steps[1].Step != "invited" { |
| 310 | t.Fatalf("the box behind the stalled read: %v", steps[1]) |
| 311 | } |
| 312 | if _, _, err := w.d.Store.Get(context.Background(), "control/host-actor-invitations/"+twoRef+".json"); err != nil { |
| 313 | t.Fatalf("the box behind the stalled read has no invitation: %v", err) |
| 314 | } |
| 315 | } |
| 316 | |
| 317 | // C2: the invitation is written only while the console's read is under a |
| 318 | // minute old. Here every look at the clock takes 31 seconds: the read is |
| 319 | // fresh when checkConfirmed has it, and over a minute old by the time the |
| 320 | // invitation would be written. That is unknown: the attempt and its wait |
| 321 | // stand. |
| 322 | func TestAnInvitationNeedsAReadUnderAMinuteOldAtTheWrite(t *testing.T) { |
| 323 | w := newEnrollWorld(t, "") |
| 324 | w.pass(t) |
| 325 | first := w.attempt(t).AttemptID |
| 326 | w.e.Now = func() time.Time { w.clock = w.clock.Add(31 * time.Second); return w.clock } |
| 327 | if got := w.pass(t); !strings.Contains(got, "more than a minute after its membership read") { |
| 328 | t.Fatalf("slow pass: %q", got) |
| 329 | } |
| 330 | if _, ok := w.invitation(t); ok { |
| 331 | t.Fatal("invited on a read over a minute old") |
| 332 | } |
| 333 | if w.attempt(t).AttemptID != first || w.e.attempts[newRef].tries != 1 || w.e.attempts[newRef].refused != "" { |
| 334 | t.Fatalf("a missed deadline counted against the attempt: %+v", w.e.attempts[newRef]) |
| 335 | } |
| 336 | } |
| 337 | |
| 338 | // slowStore is the deployment bucket with one slow read: the first Get of |
| 339 | // slow after armed is set moves the console's clock on by exactly 60 |
| 340 | // seconds, to the deadline itself, as a bucket that answers late would. It honours its context, as the S3 |
| 341 | // store does, and records every Put. |
| 342 | type slowStore struct { |
| 343 | release.Store |
| 344 | slow string |
| 345 | armed bool |
| 346 | late func() |
| 347 | puts []string |
| 348 | } |
| 349 | |
| 350 | func (s *slowStore) Get(ctx context.Context, key string) ([]byte, string, error) { |
| 351 | if err := ctx.Err(); err != nil { |
| 352 | return nil, "", err |
| 353 | } |
| 354 | if s.armed && key == s.slow { |
| 355 | s.armed = false |
| 356 | s.late() |
| 357 | } |
| 358 | return s.Store.Get(ctx, key) |
| 359 | } |
| 360 | |
| 361 | func (s *slowStore) Put(ctx context.Context, key string, body []byte, ifMatch string) (string, error) { |
| 362 | if err := ctx.Err(); err != nil { |
| 363 | return "", err |
| 364 | } |
| 365 | s.puts = append(s.puts, key) |
| 366 | return s.Store.Put(ctx, key, body, ifMatch) |
| 367 | } |
| 368 | |
| 369 | // slowAfterRead makes the store slow on key after the console's next |
| 370 | // membership read. |
| 371 | func (w *enrollWorld) slowAfterRead(key string) *slowStore { |
| 372 | ss := &slowStore{Store: w.e.Directory.Store, slow: key, late: func() { w.clock = w.clock.Add(60 * time.Second) }} |
| 373 | w.e.Directory.Store = ss |
| 374 | w.cloud.during = func() { ss.armed = true } |
| 375 | return ss |
| 376 | } |
| 377 | |
| 378 | // S1: a read inside the sealed Invite that answers late (here, its own |
| 379 | // read of the invitation key) leaves the invitation unwritten: no Put is |
| 380 | // issued after the deadline. The attempt and its wait stand, and the next |
| 381 | // pass, with a fresh read, invites. |
| 382 | func TestASlowReadInsideInviteWritesNothing(t *testing.T) { |
| 383 | w := newEnrollWorld(t, "") |
| 384 | w.pass(t) |
| 385 | first := w.attempt(t).AttemptID |
| 386 | ss := w.slowAfterRead("control/host-actor-invitations/" + newRef + ".json") |
| 387 | if got := w.pass(t); !strings.Contains(got, "more than a minute after its membership read") { |
| 388 | t.Fatalf("slow invite: %q", got) |
| 389 | } |
| 390 | if len(ss.puts) != 0 { |
| 391 | t.Fatalf("written after the deadline: %v", ss.puts) |
| 392 | } |
| 393 | if _, ok := w.invitation(t); ok || w.attempt(t).AttemptID != first || w.e.attempts[newRef].tries != 1 { |
| 394 | t.Fatalf("invited %v, attempt %+v", ok, w.e.attempts[newRef]) |
| 395 | } |
| 396 | w.cloud.during = nil |
| 397 | if got := w.pass(t); got != "invited" || w.attempt(t).AttemptID != first { |
| 398 | t.Fatalf("next pass: %q", got) |
| 399 | } |
| 400 | } |
| 401 | |
| 402 | // S1: a read inside the sealed checkMembership that answers late (the |
| 403 | // guard-ready report Replace checks before its CAS) leaves the roster |
| 404 | // unwritten. The attempt stays unconsumed, and the next pass enrolls. |
| 405 | func TestASlowReadInsideTheRosterCheckWritesNothing(t *testing.T) { |
| 406 | w := newEnrollWorld(t, "") |
| 407 | w.pass(t) |
| 408 | if got := w.pass(t); got != "invited" { |
| 409 | t.Fatal(got) |
| 410 | } |
| 411 | inv, _ := w.invitation(t) |
| 412 | w.guardReady(t, inv.Enrollment, true) |
| 413 | first, tries := w.attempt(t).AttemptID, w.e.attempts[newRef].tries |
| 414 | ss := w.slowAfterRead("status/boxes/" + newRef + ".json") |
| 415 | if got := w.pass(t); !strings.Contains(got, "more than a minute after its membership read") { |
| 416 | t.Fatalf("slow enrolment: %q", got) |
| 417 | } |
| 418 | if len(ss.puts) != 0 { |
| 419 | t.Fatalf("written after the deadline: %v", ss.puts) |
| 420 | } |
| 421 | if a := w.attempt(t); a.AttemptID != first || a.Consumed || w.e.attempts[newRef].tries != tries || w.e.attempts[newRef].refused != "" { |
| 422 | t.Fatalf("attempt %+v, seen %+v", a, w.e.attempts[newRef]) |
| 423 | } |
| 424 | if o, _ := w.d.Read(context.Background()); o.Roster.Boxes[newRef].Enrollment != "" { |
| 425 | t.Fatal("enrolled after the deadline") |
| 426 | } |
| 427 | w.cloud.during = nil |
| 428 | w.guardReady(t, inv.Enrollment, true) |
| 429 | if got := w.pass(t); !strings.HasPrefix(got, "enrolled in ie") { |
| 430 | t.Fatalf("next pass: %q", got) |
| 431 | } |
| 432 | } |
| 433 | |
| 434 | // The ruling's second test: a box cannot satisfy another box's report |
| 435 | // check. A report under newRef's key that another instance wrote is not |
| 436 | // newRef's guard-ready report, and a record under a key that is not its |
| 437 | // instance's group reference is never invited. |
| 438 | func TestABoxCannotStandInForAnother(t *testing.T) { |
| 439 | w := newEnrollWorld(t, "") |
| 440 | w.pass(t) |
| 441 | if got := w.pass(t); got != "invited" { |
| 442 | t.Fatal(got) |
| 443 | } |
| 444 | inv, _ := w.invitation(t) |
| 445 | box := status.Box{GeneratedAt: time.Now().UTC(), Metrics: &status.BoxMetrics{Enrollment: inv.Enrollment, InstanceID: "i-other", Release: "v1.0.0", ReleaseTag: "v1.0.0", |
| 446 | Platform: "linux-amd64", BinarySHA256: w.hash, MovesProtocol: 1, BootID: "boot", FloorSHA256: w.floor.Digest, FloorGuardReady: true, SystemChannel: "stable"}} |
| 447 | raw, _ := json.Marshal(box) |
| 448 | if _, err := w.d.Store.Put(context.Background(), "status/boxes/"+newRef+".json", raw, ""); err != nil { |
| 449 | t.Fatal(err) |
| 450 | } |
| 451 | if got := w.pass(t); got != "waiting for its guard-ready report" { |
| 452 | t.Fatalf("another instance's report: %q", got) |
| 453 | } |
| 454 | // A record its box published under a key of its own choosing. |
| 455 | v := newEnrollWorld(t, "") |
| 456 | _, chosen := v.signer.Record(t, "chosen", "123", "eu-west-1", "i-chosen") |
| 457 | if _, err := v.d.Store.Put(context.Background(), boxid.Key("chosen"), chosen, ""); err != nil { |
| 458 | t.Fatal(err) |
| 459 | } |
| 460 | steps, err := v.e.EnrollPass(context.Background()) |
| 461 | if err != nil { |
| 462 | t.Fatal(err) |
| 463 | } |
| 464 | for _, s := range steps { |
| 465 | if s.Box == "chosen" && !strings.Contains(s.Step, "not under its group reference") { |
| 466 | t.Fatalf("chosen key: %q", s.Step) |
| 467 | } |
| 468 | } |
| 469 | if _, _, err := v.d.Store.Get(context.Background(), "control/host-actor-invitations/chosen.json"); err == nil { |
| 470 | t.Fatal("a box with a chosen key was invited") |
| 471 | } |
| 472 | } |
| 473 | |
| 474 | // inflightStore is the bucket with one slow Put: the first Put of key |
| 475 | // after armed is set commits (when commit is set) and then, by mode, |
| 476 | // waits for its context as an S3 request cut off by the deadline does |
| 477 | // ("cancelled"), ignores its context and never answers ("ignores"), or |
| 478 | // answers success only after the console's clock has passed the deadline |
| 479 | // ("late"). In mode "delayed" it commits nothing until hang is closed, |
| 480 | // whatever its context says, then commits and reports on done: a write |
| 481 | // the console gave up on that the bucket applies later. |
| 482 | type inflightStore struct { |
| 483 | release.Store |
| 484 | key string |
| 485 | armed, commit bool |
| 486 | mode string |
| 487 | late func() |
| 488 | hang chan struct{} |
| 489 | done chan error |
| 490 | } |
| 491 | |
| 492 | func (s *inflightStore) Put(ctx context.Context, key string, body []byte, ifMatch string) (string, error) { |
| 493 | if s.armed && key == s.key { |
| 494 | s.armed = false |
| 495 | if s.mode == "delayed" { |
| 496 | <-s.hang |
| 497 | etag, err := s.Store.Put(ctx, key, body, ifMatch) |
| 498 | s.done <- err |
| 499 | return etag, err |
| 500 | } |
| 501 | etag := "" |
| 502 | if s.commit { |
| 503 | var err error |
| 504 | if etag, err = s.Store.Put(ctx, key, body, ifMatch); err != nil { |
| 505 | return "", err |
| 506 | } |
| 507 | } |
| 508 | switch s.mode { |
| 509 | case "ignores": |
| 510 | <-s.hang |
| 511 | return etag, nil |
| 512 | case "late": |
| 513 | s.late() |
| 514 | return etag, nil |
| 515 | } |
| 516 | <-ctx.Done() |
| 517 | return "", ctx.Err() |
| 518 | } |
| 519 | return s.Store.Put(ctx, key, body, ifMatch) |
| 520 | } |
| 521 | |
| 522 | // inflightAfterRead makes key's next Put slow in mode: the next |
| 523 | // membership read takes 59.95 seconds, so its write has 50 milliseconds, |
| 524 | // and a late answer comes a second after that. |
| 525 | func (w *enrollWorld) inflightAfterRead(key string, commit bool, mode string) *inflightStore { |
| 526 | is := &inflightStore{Store: w.e.Directory.Store, key: key, commit: commit, mode: mode, hang: make(chan struct{}), done: make(chan error, 1), |
| 527 | late: func() { w.clock = w.clock.Add(time.Second) }} |
| 528 | w.e.Directory.Store = is |
| 529 | w.cloud.during = func() { w.clock = w.clock.Add(59950 * time.Millisecond); is.armed = true } |
| 530 | return is |
| 531 | } |
| 532 | |
| 533 | // S1: a Put still in flight at the deadline is cancelled there, and its |
| 534 | // result is unknown: the attempt and its wait stand, and the next pass |
| 535 | // reads the bucket as it does every pass. Here the invitation did not |
| 536 | // land, then (a second time) it did, and the next pass finds it. |
| 537 | func TestAPutInFlightAtTheDeadlineIsUnknown(t *testing.T) { |
| 538 | w := newEnrollWorld(t, "") |
| 539 | w.pass(t) |
| 540 | first, tries := w.attempt(t).AttemptID, w.e.attempts[newRef].tries |
| 541 | w.inflightAfterRead("control/host-actor-invitations/"+newRef+".json", false, "cancelled") |
| 542 | if got := w.pass(t); !strings.Contains(got, "in flight at the deadline") { |
| 543 | t.Fatalf("in flight: %q", got) |
| 544 | } |
| 545 | if _, ok := w.invitation(t); ok || w.attempt(t).AttemptID != first || w.e.attempts[newRef].tries != tries || w.e.attempts[newRef].refused != "" { |
| 546 | t.Fatalf("invited %v, seen %+v", ok, w.e.attempts[newRef]) |
| 547 | } |
| 548 | w.inflightAfterRead("control/host-actor-invitations/"+newRef+".json", true, "cancelled") |
| 549 | if got := w.pass(t); !strings.Contains(got, "in flight at the deadline") { |
| 550 | t.Fatalf("in flight, landed: %q", got) |
| 551 | } |
| 552 | w.cloud.during = nil |
| 553 | if got := w.pass(t); got != "waiting for its guard-ready report" { |
| 554 | t.Fatalf("read back: %q", got) |
| 555 | } |
| 556 | } |
| 557 | |
| 558 | // U2: a store that ignores its context and never answers is cut off by |
| 559 | // the hard per-call timeout, well before the deadline: the outcome is |
| 560 | // unknown, the attempt and its wait stand, and nothing was invited. |
| 561 | func TestAStoreThatIgnoresItsContextIsCutOffAtTheHardTimeout(t *testing.T) { |
| 562 | was := storeCallTimeout |
| 563 | storeCallTimeout = 100 * time.Millisecond |
| 564 | t.Cleanup(func() { storeCallTimeout = was }) |
| 565 | w := newEnrollWorld(t, "") |
| 566 | w.pass(t) |
| 567 | first, tries := w.attempt(t).AttemptID, w.e.attempts[newRef].tries |
| 568 | is := &inflightStore{Store: w.e.Directory.Store, key: "control/host-actor-invitations/" + newRef + ".json", mode: "ignores", hang: make(chan struct{}), armed: true} |
| 569 | t.Cleanup(func() { close(is.hang) }) |
| 570 | w.e.Directory.Store = is |
| 571 | start := time.Now() |
| 572 | got := w.pass(t) |
| 573 | if !strings.Contains(got, "did not answer in time") || time.Since(start) > 5*time.Second { |
| 574 | t.Fatalf("after %s: %q", time.Since(start), got) |
| 575 | } |
| 576 | if _, ok := w.invitation(t); ok || w.attempt(t).AttemptID != first || w.e.attempts[newRef].tries != tries || w.e.attempts[newRef].refused != "" { |
| 577 | t.Fatalf("invited %v, seen %+v", ok, w.e.attempts[newRef]) |
| 578 | } |
| 579 | } |
| 580 | |
| 581 | // U1, U2: an enrolling write whose outcome is unknown is settled by the |
| 582 | // next pass's roster read, with nothing more. Not landed: the same |
| 583 | // attempt, its wait unchanged, enrols on the next pass. Landed, whether |
| 584 | // cut off in flight or answering success only after the deadline: the |
| 585 | // roster lists the box, and the pass has nothing left to do for it. |
| 586 | func TestAnUnknownEnrolmentIsSettledByTheRoster(t *testing.T) { |
| 587 | ready := func(t *testing.T) *enrollWorld { |
| 588 | w := newEnrollWorld(t, "") |
| 589 | w.pass(t) |
| 590 | if got := w.pass(t); got != "invited" { |
| 591 | t.Fatal(got) |
| 592 | } |
| 593 | inv, _ := w.invitation(t) |
| 594 | w.guardReady(t, inv.Enrollment, true) |
| 595 | return w |
| 596 | } |
| 597 | t.Run("did not land", func(t *testing.T) { |
| 598 | w := ready(t) |
| 599 | first, tries := w.attempt(t).AttemptID, w.e.attempts[newRef].tries |
| 600 | w.inflightAfterRead(Key, false, "cancelled") |
| 601 | if got := w.pass(t); !strings.Contains(got, "in flight at the deadline") { |
| 602 | t.Fatalf("in flight: %q", got) |
| 603 | } |
| 604 | if a := w.attempt(t); a.Consumed || a.AttemptID != first || w.e.attempts[newRef].tries != tries { |
| 605 | t.Fatalf("attempt %+v, seen %+v", a, w.e.attempts[newRef]) |
| 606 | } |
| 607 | w.cloud.during = nil |
| 608 | if got := w.pass(t); !strings.HasPrefix(got, "enrolled in ie") { |
| 609 | t.Fatalf("next pass: %q", got) |
| 610 | } |
| 611 | }) |
| 612 | for name, mode := range map[string]string{"landed, cut off in flight": "cancelled", "landed, answered after the deadline": "late"} { |
| 613 | t.Run(name, func(t *testing.T) { |
| 614 | w := ready(t) |
| 615 | w.inflightAfterRead(Key, true, mode) |
| 616 | want := map[string]string{"cancelled": "in flight at the deadline", "late": "answered only after the deadline"}[mode] |
| 617 | if got := w.pass(t); !strings.Contains(got, want) { |
| 618 | t.Fatalf("unknown: %q", got) |
| 619 | } |
| 620 | w.cloud.during = nil |
| 621 | calls := len(w.cloud.calls) |
| 622 | steps, err := w.e.EnrollPass(context.Background()) |
| 623 | if err != nil || len(steps) != 0 || len(w.cloud.calls) != calls { |
| 624 | t.Fatalf("settled by the roster: %+v %v, calls %v", steps, err, w.cloud.calls) |
| 625 | } |
| 626 | o, _ := w.d.Read(context.Background()) |
| 627 | if b := o.Roster.Boxes[newRef]; b.State != "active" || !o.Roster.EnrollmentAttempts[newRef].Consumed { |
| 628 | t.Fatalf("roster %+v", b) |
| 629 | } |
| 630 | }) |
| 631 | } |
| 632 | } |
| 633 | |
| 634 | // V1, the accepted exposure: the console waits for no write past its |
| 635 | // bound, but a write it gave up on can still commit. Here the enrolling |
| 636 | // Put is issued before the deadline and blocks; the pass returns unknown; |
| 637 | // the instance is detached from its group (still running); the next pass |
| 638 | // finds the box absent from the roster, and its fresh describe refuses |
| 639 | // without writing anything; then the bucket applies the delayed Put, on |
| 640 | // the roster revision it was built against, which nothing has moved. The |
| 641 | // pass after that finds the box enrolled and does nothing more. Such an |
| 642 | // enrolment rests on a membership read that was true when its write was |
| 643 | // issued: the same exposure as an instance detached after enrolment, |
| 644 | // which spec 3's regional worker reconciles. |
| 645 | func TestAWriteTheConsoleGaveUpOnCanCommitLater(t *testing.T) { |
| 646 | w := newEnrollWorld(t, "") |
| 647 | w.pass(t) |
| 648 | if got := w.pass(t); got != "invited" { |
| 649 | t.Fatal(got) |
| 650 | } |
| 651 | inv, _ := w.invitation(t) |
| 652 | w.guardReady(t, inv.Enrollment, true) |
| 653 | before, _ := w.d.Read(context.Background()) |
| 654 | is := w.inflightAfterRead(Key, false, "delayed") |
| 655 | if got := w.pass(t); !strings.Contains(got, "in flight at the deadline") { |
| 656 | t.Fatalf("unknown: %q", got) |
| 657 | } |
| 658 | // Detached: the group no longer lists the instance, which still runs. |
| 659 | w.cloud.during = nil |
| 660 | w.cloud.members["i-new"] = member("") |
| 661 | calls := len(w.cloud.calls) |
| 662 | if got := w.pass(t); !strings.Contains(got, "not confirmed") || len(w.cloud.calls) != calls+1 { |
| 663 | t.Fatalf("absent, then refused: %q, calls %v", got, w.cloud.calls) |
| 664 | } |
| 665 | if o, _ := w.d.Read(context.Background()); o.ETag != before.ETag { |
| 666 | t.Fatalf("the refusing pass wrote the roster: %s, was %s", o.ETag, before.ETag) |
| 667 | } |
| 668 | close(is.hang) |
| 669 | if err := <-is.done; err != nil { |
| 670 | t.Fatalf("the delayed write did not commit: %v", err) |
| 671 | } |
| 672 | calls = len(w.cloud.calls) |
| 673 | steps, err := w.e.EnrollPass(context.Background()) |
| 674 | if err != nil || len(steps) != 0 || len(w.cloud.calls) != calls { |
| 675 | t.Fatalf("enrolled: %+v %v, calls %v", steps, err, w.cloud.calls) |
| 676 | } |
| 677 | o, _ := w.d.Read(context.Background()) |
| 678 | if b := o.Roster.Boxes[newRef]; b.State != "active" || o.Roster.Revision != before.Roster.Revision+1 { |
| 679 | t.Fatalf("roster %+v, revision %d", b, o.Roster.Revision) |
| 680 | } |
| 681 | } |
| 682 | |
| 683 | // A refused attempt is kept, with no more reads, for its wait, then begun |
| 684 | // again; each refusal doubles the wait, up to 32 times Patience. |
| 685 | func TestARefusedAttemptIsBegunAgainWithADoublingWait(t *testing.T) { |
| 686 | w := newEnrollWorld(t, "") |
| 687 | w.cloud.members["i-new"] = member("") |
| 688 | if got := w.pass(t); got != "began an attempt" { |
| 689 | t.Fatal(got) |
| 690 | } |
| 691 | first := w.attempt(t).AttemptID |
| 692 | if got := w.pass(t); !strings.Contains(got, "not confirmed") { |
| 693 | t.Fatal(got) |
| 694 | } |
| 695 | w.clock = w.clock.Add(time.Minute + 59*time.Second) |
| 696 | if got := w.pass(t); !strings.Contains(got, "waiting before another attempt") || len(w.cloud.calls) != 1 { |
| 697 | t.Fatalf("inside the wait: %q, calls %v", got, w.cloud.calls) |
| 698 | } |
| 699 | w.clock = w.clock.Add(time.Second) |
| 700 | if got := w.pass(t); got != "began an attempt" || w.attempt(t).AttemptID == first { |
| 701 | t.Fatalf("after two minutes: %q", got) |
| 702 | } |
| 703 | if got := w.pass(t); !strings.Contains(got, "not confirmed") { |
| 704 | t.Fatal(got) |
| 705 | } |
| 706 | // The second refusal waits four minutes. |
| 707 | w.clock = w.clock.Add(3*time.Minute + 59*time.Second) |
| 708 | if got := w.pass(t); !strings.Contains(got, "waiting before another attempt") { |
| 709 | t.Fatalf("second attempt before its four minutes: %q", got) |
| 710 | } |
| 711 | w.clock = w.clock.Add(time.Second) |
| 712 | if got := w.pass(t); got != "began an attempt" { |
| 713 | t.Fatalf("second attempt after four minutes: %q", got) |
| 714 | } |
| 715 | if got := w.e.wait(10); got != 64*time.Minute { |
| 716 | t.Fatalf("the wait is not capped: %s", got) |
| 717 | } |
| 718 | } |
| 719 | |
| 720 | func TestAutomaticEnrollmentRefusesWhatItCannotProve(t *testing.T) { |
| 721 | // A box from an account or region with no group is left alone. |
| 722 | w := newEnrollWorld(t, "") |
| 723 | _, raw := boxidtest.NewSigner(t).Record(t, "elsewhere", "999", "eu-west-1", "i-x") |
| 724 | if _, err := w.d.Store.Put(context.Background(), boxid.Key("elsewhere"), raw, ""); err != nil { |
| 725 | t.Fatal(err) |
| 726 | } |
| 727 | steps, err := w.e.EnrollPass(context.Background()) |
| 728 | if err != nil { |
| 729 | t.Fatal(err) |
| 730 | } |
| 731 | for _, s := range steps { |
| 732 | if s.Box == "elsewhere" && !strings.Contains(s.Step, "no region's group") { |
| 733 | t.Fatalf("elsewhere: %q", s.Step) |
| 734 | } |
| 735 | } |
| 736 | // A record whose signature does not verify under the certificates. |
| 737 | w.e.Certs = boxidtest.NewSigner(t).Certs() |
| 738 | if got := w.pass(t); !strings.Contains(got, "does not verify") { |
| 739 | t.Fatalf("unverified record: %q", got) |
| 740 | } |
| 741 | // A deployment with no published floor enrolls nothing. |
| 742 | d, sg, _ := fixture(t) |
| 743 | unpublished := &AutoEnroller{Directory: d, Signer: sg, Namespace: "platform", Pool: "default", Now: time.Now} |
| 744 | if _, err := unpublished.EnrollPass(context.Background()); !errors.Is(err, ErrNoFloor) { |
| 745 | t.Fatalf("unpublished: %v", err) |
| 746 | } |
| 747 | } |
| 748 | |
| 749 | // S5: every membership read is logged with the box, its attempt, its |
| 750 | // instance and the request ids the read's own confirmation kept, whatever |
| 751 | // came of it: a failed read, unknown (the ids the cloud gave, "none" for a |
| 752 | // call it never answered), a refusal, and an invitation. |
| 753 | func TestEveryMembershipReadIsLoggedWithItsIds(t *testing.T) { |
| 754 | w := newEnrollWorld(t, "") |
| 755 | w.pass(t) |
| 756 | id := w.attempt(t).AttemptID |
| 757 | head := "membership read for " + newRef + " (attempt " + id + ", instance i-new, request ids " |
| 758 | w.cloud.err, w.cloud.failIDs = errors.New("Throttling: rate exceeded"), []string{"asg-req", NoRequestID} |
| 759 | w.pass(t) |
| 760 | w.cloud.err, w.cloud.failIDs = nil, nil |
| 761 | w.cloud.members["i-new"] = member("routers-uk") |
| 762 | w.pass(t) |
| 763 | if len(w.logs) != 2 || !strings.Contains(w.logs[0], head+"asg-req, none, read at ") || !strings.Contains(w.logs[0], "): unknown: "+ErrMembershipUnknown.Error()) || |
| 764 | !strings.Contains(w.logs[1], head+"asg-req, ec2-req, read at ") || !strings.Contains(w.logs[1], "): refused: "+ErrNotMember.Error()) { |
| 765 | t.Fatalf("logs %q", w.logs) |
| 766 | } |
| 767 | // A check that failed before any read was made is not a membership |
| 768 | // read, and is not logged as one. |
| 769 | w.e.logRead(EnrollmentRequest{Reference: newRef, AttemptID: id}, "i-new", MembershipEvidence{}, "invited", ErrRoster) |
| 770 | if len(w.logs) != 2 { |
| 771 | t.Fatalf("a check with no read was logged: %q", w.logs[2:]) |
| 772 | } |
| 773 | // A describe that took a minute is a refusal too. |
| 774 | x := newEnrollWorld(t, "") |
| 775 | x.pass(t) |
| 776 | x.cloud.during = func() { x.clock = x.clock.Add(time.Minute) } |
| 777 | if got := x.pass(t); !strings.Contains(got, "not confirmed") || len(x.logs) != 1 || !strings.Contains(x.logs[0], "): refused: "+ErrMembershipStale.Error()) { |
| 778 | t.Fatalf("stale: %q, logs %q", got, x.logs) |
| 779 | } |
| 780 | v := newEnrollWorld(t, "") |
| 781 | v.pass(t) |
| 782 | if got := v.pass(t); got != "invited" { |
| 783 | t.Fatal(got) |
| 784 | } |
| 785 | vhead := "membership read for " + newRef + " (attempt " + v.attempt(t).AttemptID + ", instance i-new, request ids asg-req, ec2-req, read at " |
| 786 | if len(v.logs) != 1 || !strings.Contains(v.logs[0], vhead) || !strings.HasSuffix(v.logs[0], "): invited") { |
| 787 | t.Fatalf("logs %q", v.logs) |
| 788 | } |
| 789 | } |
| 790 | |
| 791 | // conflictStore refuses the next Put of key with the store's own conflict, |
| 792 | // as a conditional write another writer got to first is refused. |
| 793 | type conflictStore struct { |
| 794 | release.Store |
| 795 | key string |
| 796 | armed bool |
| 797 | } |
| 798 | |
| 799 | func (s *conflictStore) Put(ctx context.Context, key string, body []byte, ifMatch string) (string, error) { |
| 800 | if s.armed && key == s.key { |
| 801 | s.armed = false |
| 802 | return "", release.ErrCASConflict |
| 803 | } |
| 804 | return s.Store.Put(ctx, key, body, ifMatch) |
| 805 | } |
| 806 | |
| 807 | // Only a positive answer counts against an attempt. A store that refuses |
| 808 | // the invitation's or the enrolment's write with its own error (a |
| 809 | // conflict, which a call failing after the deadline also returns) is |
| 810 | // unknown: the attempt stands, its wait does not grow, and the next pass |
| 811 | // goes on from it. |
| 812 | func TestAStoreErrorIsUnknownAndKeepsTheAttempt(t *testing.T) { |
| 813 | for _, key := range []string{"control/host-actor-invitations/" + newRef + ".json", Key} { |
| 814 | t.Run(key, func(t *testing.T) { |
| 815 | w := newEnrollWorld(t, "") |
| 816 | w.pass(t) |
| 817 | if key == Key { |
| 818 | if got := w.pass(t); got != "invited" { |
| 819 | t.Fatal(got) |
| 820 | } |
| 821 | inv, _ := w.invitation(t) |
| 822 | w.guardReady(t, inv.Enrollment, true) |
| 823 | } |
| 824 | first, tries := w.attempt(t).AttemptID, w.e.attempts[newRef].tries |
| 825 | w.e.Directory.Store = &conflictStore{Store: w.e.Directory.Store, key: key, armed: true} |
| 826 | if got := w.pass(t); !strings.Contains(got, release.ErrCASConflict.Error()) { |
| 827 | t.Fatalf("conflict: %q", got) |
| 828 | } |
| 829 | if a := w.attempt(t); a.AttemptID != first || a.Consumed || w.e.attempts[newRef].tries != tries || w.e.attempts[newRef].refused != "" { |
| 830 | t.Fatalf("attempt %+v, seen %+v", a, w.e.attempts[newRef]) |
| 831 | } |
| 832 | want := map[string]string{Key: "enrolled in ie"}[key] |
| 833 | if want == "" { |
| 834 | want = "invited" |
| 835 | } |
| 836 | if got := w.pass(t); !strings.HasPrefix(got, want) || w.attempt(t).AttemptID != first { |
| 837 | t.Fatalf("next pass: %q", got) |
| 838 | } |
| 839 | }) |
| 840 | } |
| 841 | } |
| 842 | |
| 843 | // vanishingStore answers the next Get of key, once armed, with the stored |
| 844 | // object and then removes it, as an invitation deleted just after the |
| 845 | // pass read it would be. It records each Put of key in events. |
| 846 | type vanishingStore struct { |
| 847 | release.Store |
| 848 | key string |
| 849 | armed bool |
| 850 | events *[]string |
| 851 | } |
| 852 | |
| 853 | func (s *vanishingStore) Get(ctx context.Context, key string) ([]byte, string, error) { |
| 854 | body, etag, err := s.Store.Get(ctx, key) |
| 855 | if s.armed && key == s.key && err == nil { |
| 856 | s.armed = false |
| 857 | if err := s.Store.Delete(ctx, key); err != nil { |
| 858 | return nil, "", err |
| 859 | } |
| 860 | } |
| 861 | return body, etag, err |
| 862 | } |
| 863 | |
| 864 | func (s *vanishingStore) Put(ctx context.Context, key string, body []byte, ifMatch string) (string, error) { |
| 865 | if key == s.key { |
| 866 | *s.events = append(*s.events, "put") |
| 867 | } |
| 868 | return s.Store.Put(ctx, key, body, ifMatch) |
| 869 | } |
| 870 | |
| 871 | // An invitation is never written without a membership read just before |
| 872 | // it. Here the invitation vanishes right after the pass reads it: that |
| 873 | // pass waits for the box's report and writes nothing, and the next one |
| 874 | // reads membership and only then invites again. |
| 875 | func TestAVanishedInvitationIsNeverWrittenWithoutARead(t *testing.T) { |
| 876 | w := newEnrollWorld(t, "") |
| 877 | w.pass(t) |
| 878 | var events []string |
| 879 | w.cloud.during = func() { events = append(events, "read") } |
| 880 | key := "control/host-actor-invitations/" + newRef + ".json" |
| 881 | vs := &vanishingStore{Store: w.e.Directory.Store, key: key, events: &events} |
| 882 | w.e.Directory.Store = vs |
| 883 | if got := w.pass(t); got != "invited" { |
| 884 | t.Fatal(got) |
| 885 | } |
| 886 | vs.armed = true |
| 887 | if got := w.pass(t); got != "waiting for its guard-ready report" { |
| 888 | t.Fatalf("read, then vanished: %q", got) |
| 889 | } |
| 890 | if _, ok := w.invitation(t); ok { |
| 891 | t.Fatal("the invitation was written again in the pass that read it") |
| 892 | } |
| 893 | if got := w.pass(t); got != "invited" { |
| 894 | t.Fatalf("after it vanished: %q", got) |
| 895 | } |
| 896 | if got := w.pass(t); got != "waiting for its guard-ready report" { |
| 897 | t.Fatalf("invited again: %q", got) |
| 898 | } |
| 899 | last := "" |
| 900 | for _, ev := range events { |
| 901 | if ev == "put" && last != "read" { |
| 902 | t.Fatalf("an invitation Put with no membership read just before it: %v", events) |
| 903 | } |
| 904 | last = ev |
| 905 | } |
| 906 | if fmt.Sprint(events) != "[read put read put]" { |
| 907 | t.Fatalf("events %v", events) |
| 908 | } |
| 909 | } |
| 910 | |
amadan