| 1 | package hostclaim |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "encoding/json" |
| 6 | "errors" |
| 7 | "slices" |
| 8 | "strings" |
| 9 | "testing" |
| 10 | "time" |
| 11 | |
| 12 | "github.com/carlosframework/platform/internal/release" |
| 13 | ) |
| 14 | |
| 15 | // probedDrain drives a drain of the serving fixture to a passed probe on |
| 16 | // relTarget: released, pinned, installed, probing, P admitted, the probe |
| 17 | // journaled with its time. |
| 18 | func probedDrain(t *testing.T) (func(func(*Claim) error) Observed, func() Observed) { |
| 19 | t.Helper() |
| 20 | ctx := context.Background() |
| 21 | s, cur := probedDrainIn(t) |
| 22 | apply := func(tr func(*Claim) error) Observed { |
| 23 | t.Helper() |
| 24 | next, err := Update(ctx, s, cur, tr) |
| 25 | if err != nil { |
| 26 | t.Fatal(err) |
| 27 | } |
| 28 | cur = next |
| 29 | return next |
| 30 | } |
| 31 | return apply, func() Observed { return cur } |
| 32 | } |
| 33 | |
| 34 | // probedDrainIn is probedDrain's claim and its store. |
| 35 | func probedDrainIn(t *testing.T) (release.Store, Observed) { |
| 36 | t.Helper() |
| 37 | ctx := context.Background() |
| 38 | s, o := relServing(t) |
| 39 | o, err := StartRelocation(ctx, s, o, relStart(ReasonDrain, nil), relAuthority{}) |
| 40 | if err != nil { |
| 41 | t.Fatal(err) |
| 42 | } |
| 43 | o = step(t, s, o, ack(relSource)) |
| 44 | o = step(t, s, o, ack(relTarget)) |
| 45 | o = step(t, s, o, ToReleasing("rel-1", relObserved)) |
| 46 | h := o.Claim.Holder |
| 47 | o = step(t, s, o, JournalRelease("rel-1", Release{ReleasedBy: relSource, Enrollment: relEnrollment(relSource), Generation: h.Generation, Session: h.Session, FinalTXID: 40, ManifestSeq: 7, Receipt: "control/stop/r.json", ReceiptSHA256: strings.Repeat("d", 64)})) |
| 48 | if o, err = PinAfterRelease(ctx, s, o, "rel-1", relReleases{}, nil); err != nil { |
| 49 | t.Fatal(err) |
| 50 | } |
| 51 | o = step(t, s, o, JournalPin("rel-1", relPin(o, 40, 7, relObserved))) |
| 52 | o = toProbing(t, s, o, "P1", 40) |
| 53 | o = step(t, s, o, AdmitWriter(Admission{ObservedOpID: "rel-1", ObservedPhase: PhaseRelocatingProbing, ObservedSession: o.Claim.Holder.Session, ObservedGeneration: o.Claim.Holder.Generation, |
| 54 | Box: relTarget, Mode: "cold", NewGeneration: "P1", KeepStamp: true, PredecessorCertified: true, Region: "ie", PredecessorRegion: "ie"})) |
| 55 | o = step(t, s, o, RecordProbe("rel-1", Probe{Box: relTarget, Session: o.Claim.Holder.Session, At: relObserved.Add(time.Minute)})) |
| 56 | return s, o |
| 57 | } |
| 58 | |
| 59 | // copyOf is a deep copy, for trying a transition without writing it. |
| 60 | func copyOf(t *testing.T, c *Claim) *Claim { |
| 61 | t.Helper() |
| 62 | raw, err := json.Marshal(c) |
| 63 | if err != nil { |
| 64 | t.Fatal(err) |
| 65 | } |
| 66 | var out Claim |
| 67 | if err := json.Unmarshal(raw, &out); err != nil { |
| 68 | t.Fatal(err) |
| 69 | } |
| 70 | return &out |
| 71 | } |
| 72 | |
| 73 | func stopOf(o Observed) ProbeStop { |
| 74 | return ProbeStop{Box: relTarget, Session: o.Claim.Holder.Session, Receipt: "fence-1", ReceiptSHA256: strings.Repeat("e", 64)} |
| 75 | } |
| 76 | |
| 77 | // Ruling 49 step 4: the serving CAS requires the probe's stop, journaled |
| 78 | // for exactly the passed probe's box and session; it then records the |
| 79 | // completed relocation on the serving op (ruling 56). |
| 80 | func TestCompleteRelocationRequiresTheProbeStop(t *testing.T) { |
| 81 | apply, cur := probedDrain(t) |
| 82 | o := cur() |
| 83 | var p Probe |
| 84 | if !readJournal(o.Claim.Op, "probe", &p) || p.At.IsZero() { |
| 85 | t.Fatalf("the probe has no time: %+v", p) |
| 86 | } |
| 87 | if err := CompleteRelocation("rel-1", counter())(copyOf(t, o.Claim)); !errors.Is(err, ErrRelocation) { |
| 88 | t.Fatalf("completed with the probe's session running: %v", err) |
| 89 | } |
| 90 | for name, s := range map[string]ProbeStop{ |
| 91 | "another session": {Box: relTarget, Session: o.Claim.Holder.Session + 1, Receipt: "fence-1", ReceiptSHA256: strings.Repeat("e", 64)}, |
| 92 | "another box": {Box: relSource, Session: o.Claim.Holder.Session, Receipt: "fence-1", ReceiptSHA256: strings.Repeat("e", 64)}, |
| 93 | "no receipt": {Box: relTarget, Session: o.Claim.Holder.Session, ReceiptSHA256: strings.Repeat("e", 64)}, |
| 94 | "a bad digest": {Box: relTarget, Session: o.Claim.Holder.Session, Receipt: "fence-1", ReceiptSHA256: "short"}, |
| 95 | } { |
| 96 | if err := RecordProbeStop("rel-1", s)(copyOf(t, o.Claim)); !errors.Is(err, ErrRelocation) { |
| 97 | t.Fatalf("%s: recorded: %v", name, err) |
| 98 | } |
| 99 | } |
| 100 | stop := stopOf(o) |
| 101 | o = apply(RecordProbeStop("rel-1", stop)) |
| 102 | if err := RecordProbeStop("rel-1", stop)(copyOf(t, o.Claim)); !errors.Is(err, ErrAlreadyRecorded) { |
| 103 | t.Fatalf("a replay: %v", err) |
| 104 | } |
| 105 | other := stop |
| 106 | other.Receipt = "fence-2" |
| 107 | if err := RecordProbeStop("rel-1", other)(copyOf(t, o.Claim)); !errors.Is(err, ErrRelocation) { |
| 108 | t.Fatalf("a second, different stop: %v", err) |
| 109 | } |
| 110 | o = apply(CompleteRelocation("rel-1", counter())) |
| 111 | done, ok, _ := ReadCompletedRelocation(o.Claim) |
| 112 | if !ok || done.Op != "rel-1" || done.Stop != stop || done.Stamp != o.Claim.Holder.Stamp || done.Phase != PhaseRelocatingProbing || done.Generation != "P1" { |
| 113 | t.Fatalf("completed relocation: %+v %v", done, ok) |
| 114 | } |
| 115 | if o.Claim.Op.Journal["released_move"] != "" { |
| 116 | t.Fatal("a plain relocation named a move") |
| 117 | } |
| 118 | } |
| 119 | |
| 120 | // M1: a converted landing's completion names the move on the serving op, |
| 121 | // beside the completed relocation. |
| 122 | func TestCompleteRelocationJournalsReleasedMoveForAConvertedLanding(t *testing.T) { |
| 123 | apply, cur := probedDrain(t) |
| 124 | apply(func(c *Claim) error { c.Op.Journal["conversion"] = "committed-landing"; return nil }) |
| 125 | apply(RecordProbeStop("rel-1", stopOf(cur()))) |
| 126 | o := apply(CompleteRelocation("rel-1", counter())) |
| 127 | if o.Claim.Op.Journal["released_move"] != "rel-1" { |
| 128 | t.Fatalf("serving op journal: %v", o.Claim.Op.Journal) |
| 129 | } |
| 130 | if _, ok, err := ReadCompletedRelocation(o.Claim); !ok || err != nil { |
| 131 | t.Fatal("no completed relocation beside the released move") |
| 132 | } |
| 133 | } |
| 134 | |
| 135 | // Ruling 60: installed starts its own bound. |
| 136 | func TestToInstalledAtJournalsItsStart(t *testing.T) { |
| 137 | ctx := context.Background() |
| 138 | s, o := relServing(t) |
| 139 | o, err := StartRelocation(ctx, s, o, relStart(ReasonDrain, nil), relAuthority{}) |
| 140 | if err != nil { |
| 141 | t.Fatal(err) |
| 142 | } |
| 143 | o = step(t, s, o, ack(relSource)) |
| 144 | o = step(t, s, o, ack(relTarget)) |
| 145 | o = step(t, s, o, ToReleasing("rel-1", relObserved)) |
| 146 | h := o.Claim.Holder |
| 147 | o = step(t, s, o, JournalRelease("rel-1", Release{ReleasedBy: relSource, Enrollment: relEnrollment(relSource), Generation: h.Generation, Session: h.Session, FinalTXID: 40, ManifestSeq: 7, Receipt: "control/stop/r.json", ReceiptSHA256: strings.Repeat("d", 64)})) |
| 148 | if o, err = PinAfterRelease(ctx, s, o, "rel-1", relReleases{}, nil); err != nil { |
| 149 | t.Fatal(err) |
| 150 | } |
| 151 | o = step(t, s, o, JournalPin("rel-1", relPin(o, 40, 7, relObserved))) |
| 152 | o = step(t, s, o, ToInstalling("rel-1", Target{relTarget, relEnrollment(relTarget)}, "P1")) |
| 153 | if err := ToInstalledAt("rel-1", relPrepared("P1", 40, 7), time.Time{})(copyOf(t, o.Claim)); !errors.Is(err, ErrRelocation) { |
| 154 | t.Fatalf("installed without a start: %v", err) |
| 155 | } |
| 156 | o = step(t, s, o, ToInstalledAt("rel-1", relPrepared("P1", 40, 7), relObserved)) |
| 157 | if o.Claim.Op.Phase != PhaseInstalled || o.Claim.Op.Journal["installed_at"] != relObserved.Format(time.RFC3339Nano) { |
| 158 | t.Fatalf("installed: %s %v", o.Claim.Op.Phase, o.Claim.Op.Journal) |
| 159 | } |
| 160 | } |
| 161 | |
| 162 | // Astra round 6, finding 3: A to B completes, then a successor B to C |
| 163 | // asks to start before B has discharged its probe's participation (C10). |
| 164 | // Through the real start CAS: a drain or rebalance from B is refused, with |
| 165 | // the claim unchanged, until MarkCompletedDischarged; a failover that |
| 166 | // cites B's termination fence is not, since a dead B's slot binds nothing |
| 167 | // on C. |
| 168 | func TestRelocationSuccessorWaitsForTheCompletedDischarge(t *testing.T) { |
| 169 | ctx := context.Background() |
| 170 | successor := func(reason string, fence *FenceRef) RelocationStart { |
| 171 | return RelocationStart{OpID: "rel-2", By: "worker", Now: relObserved.Add(time.Hour), Region: "ie", Reason: reason, |
| 172 | SourceBox: relTarget, SourceEnrollment: relEnrollment(relTarget), TargetBox: relSpare, TargetEnrollment: relEnrollment(relSpare), |
| 173 | DirectoryRevision: 9, AdmissionID: "adm-2", Fence: fence} |
| 174 | } |
| 175 | completed := completedDrain |
| 176 | t.Run("a drain from B waits for B's discharge", func(t *testing.T) { |
| 177 | s, o := completed(t) |
| 178 | for _, reason := range []string{ReasonDrain, ReasonRebalance} { |
| 179 | if _, err := StartRelocation(ctx, s, o, successor(reason, nil), relAuthority{}); !errors.Is(err, ErrPendingChore) { |
| 180 | t.Fatalf("%s started before the discharge: %v", reason, err) |
| 181 | } |
| 182 | } |
| 183 | if now, err := Read(ctx, s, o.Claim.Host); err != nil || now.ETag != o.ETag { |
| 184 | t.Fatalf("a refused start wrote the claim: %v", err) |
| 185 | } |
| 186 | if _, err := Update(ctx, s, o, MarkCompletedDischarged("rel-9")); !errors.Is(err, ErrRelocation) { |
| 187 | t.Fatalf("another op's discharge: %v", err) |
| 188 | } |
| 189 | o = step(t, s, o, MarkCompletedDischarged("rel-1")) |
| 190 | if _, err := Update(ctx, s, o, MarkCompletedDischarged("rel-1")); !errors.Is(err, ErrAlreadyRecorded) { |
| 191 | t.Fatalf("a repeat: %v", err) |
| 192 | } |
| 193 | if done, ok, err := ReadCompletedRelocation(o.Claim); err != nil || !ok || !done.Discharged || done.Stop != stopOf(o) { |
| 194 | t.Fatalf("the mark changed the record: %+v", done) |
| 195 | } |
| 196 | next, err := StartRelocation(ctx, s, o, successor(ReasonDrain, nil), relAuthority{}) |
| 197 | if err != nil { |
| 198 | t.Fatalf("the successor after the discharge: %v", err) |
| 199 | } |
| 200 | if next.Claim.Op.ID != "rel-2" || next.Claim.Op.Intent["source_box"] != relTarget || next.Claim.Op.Intent["target_box"] != relSpare { |
| 201 | t.Fatalf("successor op %+v", next.Claim.Op) |
| 202 | } |
| 203 | }) |
| 204 | t.Run("a failover from a fenced B does not wait", func(t *testing.T) { |
| 205 | s, o := completed(t) |
| 206 | ref := putFence(t, s, relTarget) |
| 207 | next, err := StartRelocation(ctx, s, o, successor(ReasonFailover, &ref), relAuthority{}) |
| 208 | if err != nil { |
| 209 | t.Fatalf("failover from the fenced source: %v", err) |
| 210 | } |
| 211 | if next.Claim.Op.Exec != "none" || !slices.Contains(next.Claim.Lost, relTarget) { |
| 212 | t.Fatalf("failover op %+v lost %v", next.Claim.Op, next.Claim.Lost) |
| 213 | } |
| 214 | }) |
| 215 | t.Run("no completed relocation, nothing to mark", func(t *testing.T) { |
| 216 | _, o := relServing(t) |
| 217 | if err := MarkCompletedDischarged("rel-1")(copyOf(t, o.Claim)); !errors.Is(err, ErrWrongState) { |
| 218 | t.Fatalf("%v", err) |
| 219 | } |
| 220 | }) |
| 221 | } |
| 222 | |
| 223 | // completedDrain is A to B through every real transition, completed and |
| 224 | // not yet discharged on B. |
| 225 | func completedDrain(t *testing.T) (release.Store, Observed) { |
| 226 | t.Helper() |
| 227 | s, o := probedDrainIn(t) |
| 228 | o = step(t, s, o, RecordProbeStop("rel-1", stopOf(o))) |
| 229 | o = step(t, s, o, CompleteRelocation("rel-1", counter())) |
| 230 | if done, ok, err := ReadCompletedRelocation(o.Claim); err != nil || !ok || done.Discharged || o.Claim.Op.Exec != relTarget { |
| 231 | t.Fatalf("A to B did not complete undischarged: %+v %v", done, ok) |
| 232 | } |
| 233 | return s, o |
| 234 | } |
| 235 | |
| 236 | // Astra round 7, finding 2: the discharge barrier is StartGuard's too, so |
| 237 | // a move from B (and a delete) after A to B completes is refused, with the |
| 238 | // claim unchanged, until B marks the completion discharged; then the real |
| 239 | // move start CAS succeeds. A completion box on lost holds nothing. |
| 240 | func TestMoveSuccessorWaitsForTheCompletedDischarge(t *testing.T) { |
| 241 | ctx := context.Background() |
| 242 | s, o := completedDrain(t) |
| 243 | move := BeginAdmittedMove("mv-1", "console", relTarget, relSpare, "uk", relObserved.Add(time.Hour), |
| 244 | MoveAdmission{AppID: "app-1", AdmissionID: "mv-1", SourceEnrollment: relEnrollment(relTarget), TargetEnrollment: relEnrollment(relSpare)}) |
| 245 | if _, err := Update(ctx, s, o, move); !errors.Is(err, ErrPendingChore) || !errors.Is(err, ErrRelocationFinishing) { |
| 246 | t.Fatalf("a move started before B's discharge: %v", err) |
| 247 | } |
| 248 | if _, err := Update(ctx, s, o, BeginDelete("del-1", "console", "etag-1", relObserved.Add(time.Hour), []string{"ie"}, []string{relTarget}, "")); !errors.Is(err, ErrPendingChore) { |
| 249 | t.Fatalf("a delete started before B's discharge: %v", err) |
| 250 | } |
| 251 | if now, err := Read(ctx, s, o.Claim.Host); err != nil || now.ETag != o.ETag { |
| 252 | t.Fatalf("a refused start wrote the claim: %v", err) |
| 253 | } |
| 254 | lost := copyOf(t, o.Claim) |
| 255 | lost.Lost = append(lost.Lost, relTarget) |
| 256 | if err := StartGuard(lost, []string{"uk"}, []string{relSpare}, ""); err != nil { |
| 257 | t.Fatalf("a lost completion box still holds the barrier: %v", err) |
| 258 | } |
| 259 | o = step(t, s, o, MarkCompletedDischarged("rel-1")) |
| 260 | next, err := Update(ctx, s, o, move) |
| 261 | if err != nil { |
| 262 | t.Fatalf("the move after the discharge: %v", err) |
| 263 | } |
| 264 | if next.Claim.Op.Kind != KindMoving || next.Claim.Op.ID != "mv-1" || next.Claim.Op.Exec != relTarget { |
| 265 | t.Fatalf("move op %+v", next.Claim.Op) |
| 266 | } |
| 267 | } |
| 268 | |
| 269 | // Fix round 1, I1: a completed_relocation entry that is present but cannot |
| 270 | // be read is an error, never "none". Every start from serving refuses over |
| 271 | // it (a failover citing a fence included, since which box the record binds |
| 272 | // is unknown), the discharge mark refuses, and the claim is left as it is. |
| 273 | func TestAnUnreadableCompletedRelocationRefusesEveryStart(t *testing.T) { |
| 274 | ctx := context.Background() |
| 275 | for name, raw := range map[string]string{ |
| 276 | "unparseable": `{"op":"rel-1","stop":`, |
| 277 | "no box": `{"op":"rel-1","stop":{"session":4,"receipt":"fence-1"}}`, |
| 278 | "no session": `{"op":"rel-1","stop":{"box":"` + relTarget + `","receipt":"fence-1"}}`, |
| 279 | "no op": `{"stop":{"box":"` + relTarget + `","session":4,"receipt":"fence-1"}}`, |
| 280 | "empty": ``, |
| 281 | "another shape": `["rel-1"]`, |
| 282 | } { |
| 283 | t.Run(name, func(t *testing.T) { |
| 284 | s, o := completedDrain(t) |
| 285 | o = step(t, s, o, func(c *Claim) error { c.Op.Journal[JournalCompleted] = raw; return nil }) |
| 286 | if _, ok, err := ReadCompletedRelocation(o.Claim); ok || !errors.Is(err, ErrRelocation) { |
| 287 | t.Fatalf("read as %v, %v", ok, err) |
| 288 | } |
| 289 | successor := RelocationStart{OpID: "rel-2", By: "worker", Now: relObserved.Add(time.Hour), Region: "ie", Reason: ReasonDrain, |
| 290 | SourceBox: relTarget, SourceEnrollment: relEnrollment(relTarget), TargetBox: relSpare, TargetEnrollment: relEnrollment(relSpare), |
| 291 | DirectoryRevision: 9, AdmissionID: "adm-2"} |
| 292 | if _, err := StartRelocation(ctx, s, o, successor, relAuthority{}); !errors.Is(err, ErrRelocation) { |
| 293 | t.Fatalf("a drain started over an unreadable completion: %v", err) |
| 294 | } |
| 295 | ref := putFence(t, s, relTarget) |
| 296 | failover := successor |
| 297 | failover.Reason, failover.Fence = ReasonFailover, &ref |
| 298 | if _, err := StartRelocation(ctx, s, o, failover, relAuthority{}); !errors.Is(err, ErrRelocation) { |
| 299 | t.Fatalf("a failover started over an unreadable completion: %v", err) |
| 300 | } |
| 301 | if err := StartGuard(copyOf(t, o.Claim), []string{"uk"}, []string{relSpare}, ""); !errors.Is(err, ErrRelocation) { |
| 302 | t.Fatalf("StartGuard: %v", err) |
| 303 | } |
| 304 | if err := MarkCompletedDischarged("rel-1")(copyOf(t, o.Claim)); !errors.Is(err, ErrRelocation) { |
| 305 | t.Fatalf("the discharge mark over an unreadable completion: %v", err) |
| 306 | } |
| 307 | if now, err := Read(ctx, s, o.Claim.Host); err != nil || now.ETag != o.ETag { |
| 308 | t.Fatalf("a refused start wrote the claim: %v", err) |
| 309 | } |
| 310 | }) |
| 311 | } |
| 312 | } |
| 313 | |
amadan