| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "errors" |
| 6 | "fmt" |
| 7 | "log" |
| 8 | "sort" |
| 9 | "strings" |
| 10 | "time" |
| 11 | |
| 12 | "github.com/carlosframework/platform/internal/hostbox" |
| 13 | ) |
| 14 | |
| 15 | const servingParkRecoveryLimit = 128 |
| 16 | |
| 17 | var errServingParkRetagPending = errors.New("the earlier park is incomplete; retain its evidence and inspect the root service journal before retrying") |
| 18 | |
| 19 | type servingParkRecoveryResult struct { |
| 20 | Host, ID string |
| 21 | Err error |
| 22 | } |
| 23 | |
| 24 | type servingParkRecoveryPage struct { |
| 25 | Results []servingParkRecoveryResult |
| 26 | Next string |
| 27 | More bool |
| 28 | } |
| 29 | |
| 30 | // The cursor bounds record processing and lets a later pass reach records |
| 31 | // beyond a blocked host. Store.Names still enumerates the private directory. |
| 32 | func (l *servingParkLedger) recoveryIDs(ctx context.Context, after string, limit int) ([]string, bool, error) { |
| 33 | if ctx == nil || l == nil || l.state == nil || limit < 1 || limit > servingParkRecoveryLimit { |
| 34 | return nil, false, hostbox.ErrUnsupported |
| 35 | } |
| 36 | if err := ctx.Err(); err != nil { |
| 37 | return nil, false, err |
| 38 | } |
| 39 | names, err := l.state.Names() |
| 40 | if err != nil { |
| 41 | return nil, false, err |
| 42 | } |
| 43 | var ids []string |
| 44 | for _, name := range names { |
| 45 | if strings.HasPrefix(name, "retirement-") && strings.HasSuffix(name, ".json") { |
| 46 | id := strings.TrimSuffix(strings.TrimPrefix(name, "retirement-"), ".json") |
| 47 | if id > after { |
| 48 | ids = append(ids, id) |
| 49 | } |
| 50 | } |
| 51 | } |
| 52 | sort.Strings(ids) |
| 53 | more := len(ids) > limit |
| 54 | if more { |
| 55 | ids = ids[:limit] |
| 56 | } |
| 57 | return ids, more, ctx.Err() |
| 58 | } |
| 59 | |
| 60 | // Completed deletion needs no live claim or agent. Its original durable cause |
| 61 | // and recorded identities remain the authority after a host is removed. |
| 62 | func (l *servingParkLedger) recoverCompleted(ctx context.Context, id string) error { |
| 63 | return l.withNamespace(ctx, func() error { |
| 64 | r, err := l.read(id) |
| 65 | if err != nil { |
| 66 | return err |
| 67 | } |
| 68 | unlock, err := l.state.Lock(ctx, "slot-"+servingParkKey(r.Authorization.Successor.Box, r.Authorization.Successor.Host)+".lock") |
| 69 | if err != nil { |
| 70 | return err |
| 71 | } |
| 72 | defer unlock() |
| 73 | unlockRecord, err := l.state.Lock(ctx, "retirement-"+id+".lock") |
| 74 | if err != nil { |
| 75 | return err |
| 76 | } |
| 77 | defer unlockRecord() |
| 78 | if !r.Complete { |
| 79 | return hostbox.ErrChanged |
| 80 | } |
| 81 | if r.Deleting { |
| 82 | if !r.Closed { |
| 83 | err = l.deleteLocked(ctx, &r, r.Cause) |
| 84 | } else { |
| 85 | // Older records can be closed with their empty event retained. |
| 86 | err = l.fs.Delete(&r, func() error { |
| 87 | if err := ctx.Err(); err != nil { |
| 88 | return err |
| 89 | } |
| 90 | return l.save(r) |
| 91 | }) |
| 92 | } |
| 93 | if err != nil { |
| 94 | return err |
| 95 | } |
| 96 | } |
| 97 | // Completion may be durable while clearing Pending lost its reply. |
| 98 | // Last and ID are replay references and survive this cleanup. |
| 99 | slot, err := l.slot(r.Authorization) |
| 100 | if err != nil { |
| 101 | return err |
| 102 | } |
| 103 | if slot.Pending == id { |
| 104 | if err = ctx.Err(); err != nil { |
| 105 | return err |
| 106 | } |
| 107 | slot.Pending = "" |
| 108 | if err = l.write("slot-"+servingParkKey(slot.Box, slot.Host)+".json", slot); err != nil { |
| 109 | return err |
| 110 | } |
| 111 | } |
| 112 | return l.collectClosedLocked(ctx, &r, slot) |
| 113 | }) |
| 114 | } |
| 115 | |
| 116 | func (p *servingParkRuntime) recoverPending(ctx context.Context) error { |
| 117 | if p == nil { |
| 118 | return nil |
| 119 | } |
| 120 | call, done, err := p.begin(ctx) |
| 121 | if err != nil { |
| 122 | return err |
| 123 | } |
| 124 | defer done() |
| 125 | recovery, ok := p.authority.(interface { |
| 126 | recoverServingParks(context.Context, string, int) (servingParkRecoveryPage, error) |
| 127 | }) |
| 128 | if !ok { |
| 129 | return hostbox.ErrUnsupported |
| 130 | } |
| 131 | p.recoveryMu.Lock() |
| 132 | defer p.recoveryMu.Unlock() |
| 133 | page, err := recovery.recoverServingParks(call, p.recoveryAfter, servingParkRecoveryLimit) |
| 134 | if err == nil { |
| 135 | p.recoveryAfter = "" |
| 136 | if page.More { |
| 137 | p.recoveryAfter = page.Next |
| 138 | } |
| 139 | } |
| 140 | var failures []error |
| 141 | if err != nil { |
| 142 | failures = append(failures, hostbox.ErrChanged) |
| 143 | } |
| 144 | for _, result := range page.Results { |
| 145 | if result.Err != nil { |
| 146 | // Draft retirement.reap_failed and reason.park_pending from the index. |
| 147 | reason := "a local park operation failed; inspect the root service journal" |
| 148 | if errors.Is(result.Err, errServingParkRetagPending) { |
| 149 | reason = errServingParkRetagPending.Error() |
| 150 | } else if errors.Is(result.Err, hostbox.ErrChanged) || errors.Is(result.Err, ErrClaimStopReceipt) { |
| 151 | reason = "the parked artifact or its parent no longer matches the root-recorded identity" |
| 152 | } |
| 153 | if result.Host == "unknown" { |
| 154 | reason = "the private retirement record could not be validated; retain the data for investigation" |
| 155 | } |
| 156 | failures = append(failures, fmt.Errorf("Host %q: serving database retirement %q retained: %s.", result.Host, result.ID, reason)) |
| 157 | } |
| 158 | } |
| 159 | if len(failures) > 0 && time.Since(p.lastRecoveryLog) >= time.Hour { |
| 160 | log.Print(errors.Join(failures...)) |
| 161 | p.lastRecoveryLog = time.Now() |
| 162 | } |
| 163 | return errors.Join(err, call.Err()) |
| 164 | } |
| 165 | |
amadan