| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "errors" |
| 6 | "log" |
| 7 | "sort" |
| 8 | "time" |
| 9 | |
| 10 | "github.com/carlosframework/platform/internal/hostbox" |
| 11 | "github.com/carlosframework/platform/internal/instancectl" |
| 12 | ) |
| 13 | |
| 14 | // Select at most one page of the oldest complete parks. Incomplete records and |
| 15 | // unknown contents never contribute reclaimable space. The namespace lock |
| 16 | // serializes selection, slot ownership and identity-checked deletion. |
| 17 | func (l *servingParkLedger) maintainParks(ctx context.Context, room func() error) error { |
| 18 | if room == nil { |
| 19 | return hostbox.ErrUnsupported |
| 20 | } |
| 21 | return l.withNamespace(ctx, func() error { |
| 22 | if err := ctx.Err(); err != nil { |
| 23 | return err |
| 24 | } |
| 25 | if err := l.collectOrphanBindings(ctx); err != nil { |
| 26 | return err |
| 27 | } |
| 28 | pressure := room() |
| 29 | if pressure != nil && !errors.Is(pressure, instancectl.ErrDiskFull) { |
| 30 | return pressure |
| 31 | } |
| 32 | var candidates []servingParkRecord |
| 33 | var failures []error |
| 34 | after := "" |
| 35 | for { |
| 36 | ids, more, err := l.recoveryIDs(ctx, after, servingParkRecoveryLimit) |
| 37 | if err != nil { |
| 38 | return err |
| 39 | } |
| 40 | for _, id := range ids { |
| 41 | if err := ctx.Err(); err != nil { |
| 42 | return err |
| 43 | } |
| 44 | r, err := l.read(id) |
| 45 | if err != nil { |
| 46 | if len(failures) < servingParkRecoveryLimit { |
| 47 | failures = append(failures, err) |
| 48 | } |
| 49 | continue |
| 50 | } |
| 51 | if r.Complete && !r.Closed && r.Disposition == servingParkRetain && (pressure != nil || !l.now().Before(r.ReapAfter)) { |
| 52 | candidates = append(candidates, r) |
| 53 | sort.Slice(candidates, func(i, j int) bool { |
| 54 | if candidates[i].ParkedAt.Equal(candidates[j].ParkedAt) { |
| 55 | return candidates[i].Authorization.ID < candidates[j].Authorization.ID |
| 56 | } |
| 57 | return candidates[i].ParkedAt.Before(candidates[j].ParkedAt) |
| 58 | }) |
| 59 | if len(candidates) > servingParkRecoveryLimit { |
| 60 | candidates = candidates[:servingParkRecoveryLimit] |
| 61 | } |
| 62 | } |
| 63 | } |
| 64 | if !more { |
| 65 | break |
| 66 | } |
| 67 | after = ids[len(ids)-1] |
| 68 | } |
| 69 | for _, r := range candidates { |
| 70 | if err := ctx.Err(); err != nil { |
| 71 | return err |
| 72 | } |
| 73 | cause := "expiry" |
| 74 | if l.now().Before(r.ReapAfter) { |
| 75 | err := room() |
| 76 | if err == nil { |
| 77 | continue |
| 78 | } |
| 79 | if !errors.Is(err, instancectl.ErrDiskFull) { |
| 80 | return err |
| 81 | } |
| 82 | cause = "pressure" |
| 83 | } |
| 84 | err := func() error { |
| 85 | unlock, err := l.state.Lock(ctx, "slot-"+servingParkKey(r.Authorization.Successor.Box, r.Authorization.Successor.Host)+".lock") |
| 86 | if err != nil { |
| 87 | return err |
| 88 | } |
| 89 | defer unlock() |
| 90 | unlock, err = l.state.Lock(ctx, "retirement-"+r.Authorization.ID+".lock") |
| 91 | if err != nil { |
| 92 | return err |
| 93 | } |
| 94 | defer unlock() |
| 95 | if r.Deleting { |
| 96 | cause = r.Cause |
| 97 | } |
| 98 | if err := l.deleteLocked(ctx, &r, cause); err != nil { |
| 99 | return err |
| 100 | } |
| 101 | slot, err := l.slot(r.Authorization) |
| 102 | if err != nil { |
| 103 | return err |
| 104 | } |
| 105 | return l.collectClosedLocked(ctx, &r, slot) |
| 106 | }() |
| 107 | if err != nil { |
| 108 | failures = append(failures, err) |
| 109 | } |
| 110 | } |
| 111 | return errors.Join(failures...) |
| 112 | }) |
| 113 | } |
| 114 | |
| 115 | func (p *servingParkRuntime) Start(ctx context.Context) { |
| 116 | if p == nil { |
| 117 | return |
| 118 | } |
| 119 | p.mu.Lock() |
| 120 | if p.started || p.closing { |
| 121 | p.mu.Unlock() |
| 122 | return |
| 123 | } |
| 124 | p.started = true |
| 125 | p.mu.Unlock() |
| 126 | call, done, err := p.begin(ctx) |
| 127 | if err != nil { |
| 128 | return |
| 129 | } |
| 130 | go func() { |
| 131 | defer done() |
| 132 | ticker := time.NewTicker(time.Minute) |
| 133 | defer ticker.Stop() |
| 134 | after := "" |
| 135 | lastLog := time.Time{} |
| 136 | for { |
| 137 | bounded, cancel := context.WithCancel(call) |
| 138 | if a, ok := p.authority.(interface { |
| 139 | maintainServingParks(context.Context, string) (string, error) |
| 140 | }); ok { |
| 141 | var err error |
| 142 | after, err = a.maintainServingParks(bounded, after) |
| 143 | if err != nil && call.Err() == nil && time.Since(lastLog) >= time.Hour { |
| 144 | lastLog = time.Now() |
| 145 | log.Print(err) |
| 146 | } |
| 147 | } |
| 148 | cancel() |
| 149 | select { |
| 150 | case <-call.Done(): |
| 151 | return |
| 152 | case <-ticker.C: |
| 153 | } |
| 154 | } |
| 155 | }() |
| 156 | } |
| 157 | |
| 158 | func maintainServingParkPass(ctx context.Context, after string, cleanup func(context.Context) error, recover func(context.Context, string, int) (servingParkRecoveryPage, error), budget time.Duration) (string, error) { |
| 159 | pressure, stopPressure := context.WithTimeout(ctx, budget) |
| 160 | pressureErr := cleanup(pressure) |
| 161 | stopPressure() |
| 162 | recovery, stopRecovery := context.WithTimeout(ctx, budget) |
| 163 | defer stopRecovery() |
| 164 | page, err := recover(recovery, after, servingParkRecoveryLimit) |
| 165 | for _, result := range page.Results { |
| 166 | err = errors.Join(err, result.Err) |
| 167 | } |
| 168 | next := "" |
| 169 | if page.More { |
| 170 | next = page.Next |
| 171 | } |
| 172 | return next, errors.Join(pressureErr, err) |
| 173 | } |
| 174 | |
amadan