| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "encoding/json" |
| 5 | "errors" |
| 6 | "os" |
| 7 | "os/exec" |
| 8 | "path/filepath" |
| 9 | "sync" |
| 10 | "testing" |
| 11 | "time" |
| 12 | |
| 13 | "github.com/carlosframework/platform/internal/hostwrite" |
| 14 | "github.com/carlosframework/platform/internal/provision" |
| 15 | "github.com/carlosframework/platform/internal/registry" |
| 16 | "github.com/carlosframework/platform/internal/release" |
| 17 | "github.com/carlosframework/platform/internal/rootstate" |
| 18 | ) |
| 19 | |
| 20 | func TestPlan14HistoryRecoveryProcessCrash(t *testing.T) { |
| 21 | t.Parallel() |
| 22 | if boundary := os.Getenv("PLAN14_HISTORY_CRASH"); boundary != "" { |
| 23 | dir := os.Getenv("PLAN14_HISTORY_DIR") |
| 24 | e, rt, u, c, pub := plan14AdoptionFixture(t) |
| 25 | storeDir := filepath.Join(dir, "deployment") |
| 26 | rec, _, err := provision.FindRecordByHost(t.Context(), e.store, rt.Host) |
| 27 | if err != nil { |
| 28 | t.Fatal(err) |
| 29 | } |
| 30 | plan14PutRecord(t, release.NewDirStore(storeDir), *rec) |
| 31 | h := seedHostRootHistory(t, rt) |
| 32 | disk := &rollbackDisk{dir: filepath.Join(dir, "attempts")} |
| 33 | historyDisk := &rollbackDisk{dir: filepath.Join(dir, "history")} |
| 34 | for _, d := range []*rollbackDisk{disk, historyDisk} { |
| 35 | if err := os.Mkdir(d.dir, 0700); err != nil { |
| 36 | t.Fatal(err) |
| 37 | } |
| 38 | } |
| 39 | for k, raw := range h.Store.(*rollbackMemory).files { |
| 40 | if err := historyDisk.Write(k, raw); err != nil { |
| 41 | t.Fatal(err) |
| 42 | } |
| 43 | } |
| 44 | h.Store = historyDisk |
| 45 | deps := boundRootHistory(h, c.resolve, e.root.obs) |
| 46 | c.attempts = func() (rootstate.Store, error) { return disk, nil } |
| 47 | c.history = func() (*release.RootHistoryPass, error) { return deps, nil } |
| 48 | linkRootHistory(deps, c, u.opt, func(string) string { return "5s" }) |
| 49 | crash := func() { |
| 50 | raw, err := json.Marshal(rollbackProcessFixture{Props: u.props, Exe: u.exe, Opt: u.opt, Bake: e.bake, Store: storeDir, Effects: u.effects}) |
| 51 | if err != nil { |
| 52 | t.Fatal(err) |
| 53 | } |
| 54 | if err := os.WriteFile(filepath.Join(dir, "fixture.json"), raw, 0600); err != nil { |
| 55 | t.Fatal(err) |
| 56 | } |
| 57 | os.Exit(74) |
| 58 | } |
| 59 | disk.afterWrite = func(k string, raw []byte) { |
| 60 | if k != rollbackJournalKey(rt.Host) { |
| 61 | return |
| 62 | } |
| 63 | var a rollbackAttempt |
| 64 | if json.Unmarshal(raw, &a) != nil || a.History == nil { |
| 65 | return |
| 66 | } |
| 67 | if boundary == "linked" && !a.Failed && !a.Recovering && !a.Settled { |
| 68 | crash() |
| 69 | } |
| 70 | if boundary == "forward settled" && a.Settled { |
| 71 | crash() |
| 72 | } |
| 73 | if boundary == "rollback pending" && a.Recovering && !a.Settled { |
| 74 | crash() |
| 75 | } |
| 76 | } |
| 77 | historyDisk.afterWrite = func(k string, raw []byte) { |
| 78 | if boundary != "history settled" { |
| 79 | return |
| 80 | } |
| 81 | var r struct { |
| 82 | Binding release.HistoryBinding `json:"binding"` |
| 83 | Attempt *release.HistoryAttempt `json:"attempt"` |
| 84 | } |
| 85 | if json.Unmarshal(raw, &r) == nil && r.Binding.Host == rt.Host && r.Attempt == nil && len(u.effects) > 0 { |
| 86 | crash() |
| 87 | } |
| 88 | } |
| 89 | u.crashNew = boundary != "forward running" && boundary != "forward unhealthy" && boundary != "forward settled" && boundary != "history settled" |
| 90 | u.afterEffect = func() { |
| 91 | if (boundary == "forward failed" || boundary == "forward running" || boundary == "forward unhealthy" || boundary == "rollback fails") && len(u.effects) == 1 { |
| 92 | crash() |
| 93 | } |
| 94 | if boundary == "rollback running" && len(u.effects) == 3 { |
| 95 | crash() |
| 96 | } |
| 97 | } |
| 98 | forward := rootRestartRouteWith(e.root.reg, u, u.opt, func(string) string { return "5s" }, nil, c) |
| 99 | _ = release.Pass(t.Context(), e.root.reg, e.store, pub, u.opt, forward, release.WithRootHistory(deps), release.WithRestartFailurePropagation(), release.WithRollbackRestart(rollbackUnitRestart(e.root.reg, u, u.opt, func(string) string { return "5s" }, c))) |
| 100 | t.Fatal("crash boundary not reached") |
| 101 | } |
| 102 | for _, boundary := range []string{"linked", "forward failed", "forward running", "forward unhealthy", "forward settled", "rollback pending", "rollback running", "history settled", "rollback fails"} { |
| 103 | t.Run(boundary, func(t *testing.T) { |
| 104 | dir := t.TempDir() |
| 105 | cmd := exec.CommandContext(t.Context(), os.Args[0], "-test.run=^TestPlan14HistoryRecoveryProcessCrash$", "-test.count=1") |
| 106 | cmd.Dir = dir |
| 107 | cmd.Env = append(os.Environ(), "TMPDIR="+dir, "PLAN14_HISTORY_CRASH="+boundary, "PLAN14_HISTORY_DIR="+dir, "CARLOS_HOST_CLAIM=") |
| 108 | out, err := cmd.CombinedOutput() |
| 109 | var exit *exec.ExitError |
| 110 | if !errors.As(err, &exit) || exit.ExitCode() != 74 { |
| 111 | t.Fatalf("child: %v\n%s", err, out) |
| 112 | } |
| 113 | raw, err := os.ReadFile(filepath.Join(dir, "fixture.json")) |
| 114 | if err != nil { |
| 115 | t.Fatal(err) |
| 116 | } |
| 117 | var f rollbackProcessFixture |
| 118 | if err := json.Unmarshal(raw, &f); err != nil { |
| 119 | t.Fatal(err) |
| 120 | } |
| 121 | u := &plan14RestartUnits{props: f.Props, exe: f.Exe, opt: f.Opt, effects: f.Effects, badRollback: boundary == "rollback fails"} |
| 122 | disk := &rollbackDisk{dir: filepath.Join(dir, "attempts")} |
| 123 | source := &proofBindingSource{store: release.NewDirStore(f.Store), sys: u, optDir: u.opt, openFile: os.Open, unitDirs: []string{filepath.Dir(u.props.FragmentPath)}, healthy: func(acct, app, ver, digest string) *release.HealthyBuild { |
| 124 | for _, b := range release.ReadHealthy(f.Bake, acct, app) { |
| 125 | if b.Version == ver && b.SHA256 == digest { |
| 126 | return &b |
| 127 | } |
| 128 | } |
| 129 | return nil |
| 130 | }} |
| 131 | a, err := readRollbackAttempt(disk, observedHost) |
| 132 | if err != nil || a.History == nil { |
| 133 | t.Fatalf("missing linked intent: %+v %v", a, err) |
| 134 | } |
| 135 | now := a.Created.Add(time.Second) |
| 136 | ready := func(registry.Route) func() error { |
| 137 | return func() error { |
| 138 | if boundary == "forward unhealthy" && u.exe == release.VersionPath(u.opt, "bab", "v2", "console") { |
| 139 | return os.ErrPermission |
| 140 | } |
| 141 | return nil |
| 142 | } |
| 143 | } |
| 144 | h := &release.RootHistory{Store: &rollbackDisk{dir: filepath.Join(dir, "history")}} |
| 145 | obs := proofObserver{resolve: source.resolve, sys: u, optDir: u.opt, procExe: func(int) string { return u.exe }, now: time.Now, ready: ready} |
| 146 | deps := boundRootHistory(h, source.resolve, obs) |
| 147 | c := rootRestartChecks{history: func() (*release.RootHistoryPass, error) { return deps, nil }, started: &sync.Map{}, attempts: func() (rootstate.Store, error) { return disk, nil }, prepareRollback: source.prepareRollback, rollback: source.rollback, host: source.forHost, ready: ready, now: func() time.Time { return now }, sleep: func(d time.Duration) { now = now.Add(d) }} |
| 148 | before := len(u.effects) |
| 149 | err = recoverRootRestarts(t.Context(), u, u.opt, func(string) string { return "5s" }, c, hostwrite.Config{}) |
| 150 | refs, refsErr := h.Refusals(a.HistoryBinding) |
| 151 | if boundary == "rollback fails" { |
| 152 | if err == nil || refsErr == nil { |
| 153 | t.Fatalf("failed rollback settled: %v %v", err, refsErr) |
| 154 | } |
| 155 | before := len(u.effects) |
| 156 | if err := recoverRootRestarts(t.Context(), u, u.opt, func(string) string { return "5s" }, c, hostwrite.Config{}); err != nil || len(u.effects) != before { |
| 157 | t.Fatalf("terminal rollback repeated: %v %v", err, u.effects) |
| 158 | } |
| 159 | if _, err := h.Refusals(a.HistoryBinding); err == nil { |
| 160 | t.Fatal("terminal failure erased history intent") |
| 161 | } |
| 162 | return |
| 163 | } |
| 164 | if err != nil || refsErr != nil { |
| 165 | t.Fatalf("resume: %v %v", err, refsErr) |
| 166 | } |
| 167 | forward := boundary == "forward running" || boundary == "forward settled" || boundary == "history settled" |
| 168 | if forward && (len(refs) != 0 || u.exe != release.VersionPath(u.opt, "bab", "v2", "console")) { |
| 169 | t.Fatalf("forward not recovered: %+v %s", refs, u.exe) |
| 170 | } |
| 171 | if !forward && (len(refs) != 1 || !refs[0].Recovered || refs[0].Version != "v2" || refs[0].RestoredVersion != "v1") { |
| 172 | t.Fatalf("rollback deny missing: %+v", refs) |
| 173 | } |
| 174 | if boundary != "forward failed" && boundary != "forward unhealthy" && boundary != "rollback pending" && len(u.effects) != before { |
| 175 | t.Fatal("recovery repeated an observed effect") |
| 176 | } |
| 177 | before = len(u.effects) |
| 178 | if err := recoverRootRestarts(t.Context(), u, u.opt, func(string) string { return "5s" }, c, hostwrite.Config{}); err != nil || len(u.effects) != before { |
| 179 | t.Fatalf("replay: %v", err) |
| 180 | } |
| 181 | }) |
| 182 | } |
| 183 | } |
| 184 | |