| 1 | // This file is apps/<acct>/<app>/schedule-status/<host>.json (spec §4.3): |
| 2 | // the box's best-effort report of what each of its instances will do next |
| 3 | // and what happened last time, written after every fire and every converge |
| 4 | // that changed a host's rows. It is what makes `carlos schedule ls` honest |
| 5 | // for a member with no box access — the declaration alone says what was |
| 6 | // asked; these objects say what is actually going to happen. |
| 7 | |
| 8 | package schedule |
| 9 | |
| 10 | import ( |
| 11 | "context" |
| 12 | "encoding/json" |
| 13 | "errors" |
| 14 | "fmt" |
| 15 | "os" |
| 16 | "sort" |
| 17 | "strings" |
| 18 | |
| 19 | "github.com/carlosframework/platform/internal/release" |
| 20 | ) |
| 21 | |
| 22 | // TimerStatus is one row of a box's report, and one row of `carlos schedule |
| 23 | // ls` — the same shape end to end so the member reads what the box knows. |
| 24 | type TimerStatus struct { |
| 25 | Name string `json:"name"` |
| 26 | Kind string `json:"kind"` |
| 27 | Spec string `json:"spec"` |
| 28 | Path string `json:"path"` |
| 29 | NextAt int64 `json:"next_at"` |
| 30 | DueAt int64 `json:"due_at"` // the occurrence next_at delivers (X-Carlos-Schedule-At); 0 from a pre-due_at box |
| 31 | LastAt int64 `json:"last_at"` |
| 32 | LastStatus string `json:"last_status"` |
| 33 | LastMs int64 `json:"last_ms"` |
| 34 | Attempts int `json:"attempts"` |
| 35 | |
| 36 | // HoldsAwake is the box's warning: this row fires at least as often as |
| 37 | // the deployment's idle window, so its instance never hibernates between |
| 38 | // ticks and accrues instance-hours around the clock. Computed by the box, |
| 39 | // which is the only place that holds both numbers. False from a box that |
| 40 | // predates the field, and false wherever hibernation is not armed. |
| 41 | HoldsAwake bool `json:"holds_awake,omitempty"` |
| 42 | } |
| 43 | |
| 44 | // Status is apps/<acct>/<app>/schedule-status/<host>.json (spec §4.3): |
| 45 | // what THIS instance will do next and what happened last, written by the |
| 46 | // box that holds it. Unsupported, when set, says why Timers is empty. |
| 47 | type Status struct { |
| 48 | Schema int `json:"schema"` |
| 49 | Host string `json:"host"` |
| 50 | Box string `json:"box"` |
| 51 | ReportedAt string `json:"reported_at"` |
| 52 | Unsupported string `json:"unsupported"` |
| 53 | Timers []TimerStatus `json:"timers"` |
| 54 | } |
| 55 | |
| 56 | // WriteStatus replaces the host's report. The store's Put is CAS-only, so |
| 57 | // this reads the current etag first; a conflict (two writers for one host — |
| 58 | // not a designed state) re-reads once. Never loops: a report is |
| 59 | // best-effort and the next fire rewrites it anyway. |
| 60 | func claimRawWriteStatus(ctx context.Context, s release.Store, acct, app string, st Status) error { |
| 61 | if err := checkIDs(acct, app); err != nil { |
| 62 | return err |
| 63 | } |
| 64 | st.Schema = 1 |
| 65 | if st.Timers == nil { |
| 66 | st.Timers = []TimerStatus{} |
| 67 | } |
| 68 | raw, err := json.Marshal(st) |
| 69 | if err != nil { |
| 70 | return err |
| 71 | } |
| 72 | key := StatusKey(acct, app, st.Host) |
| 73 | for attempt := 0; attempt < 2; attempt++ { |
| 74 | _, etag, err := s.Get(ctx, key) |
| 75 | if errors.Is(err, os.ErrNotExist) { |
| 76 | etag = "" |
| 77 | } else if err != nil { |
| 78 | return err |
| 79 | } |
| 80 | if _, err := s.Put(ctx, key, raw, etag); err == nil { |
| 81 | return nil |
| 82 | } else if !errors.Is(err, release.ErrCASConflict) { |
| 83 | return err |
| 84 | } |
| 85 | } |
| 86 | return fmt.Errorf("status for %s: conflicting writer", st.Host) |
| 87 | } |
| 88 | |
| 89 | // ListStatus reads every host's report for the app, sorted by host. An |
| 90 | // unreadable object is skipped and its error joined into the return, with |
| 91 | // the readable ones still returned — one bad report must not blank the list. |
| 92 | func ListStatus(ctx context.Context, s release.Store, acct, app string) ([]Status, error) { |
| 93 | if err := checkIDs(acct, app); err != nil { |
| 94 | return nil, err |
| 95 | } |
| 96 | keys, err := s.List(ctx, StatusPrefix(acct, app)) |
| 97 | if err != nil { |
| 98 | return nil, err |
| 99 | } |
| 100 | var out []Status |
| 101 | var errs []error |
| 102 | for _, k := range keys { |
| 103 | if !strings.HasSuffix(k, ".json") { |
| 104 | continue |
| 105 | } |
| 106 | raw, _, err := s.Get(ctx, k) |
| 107 | if err != nil { |
| 108 | errs = append(errs, err) |
| 109 | continue |
| 110 | } |
| 111 | var st Status |
| 112 | if err := json.Unmarshal(raw, &st); err != nil { |
| 113 | errs = append(errs, fmt.Errorf("%s: %w", k, err)) |
| 114 | continue |
| 115 | } |
| 116 | if st.Timers == nil { |
| 117 | st.Timers = []TimerStatus{} |
| 118 | } |
| 119 | out = append(out, st) |
| 120 | } |
| 121 | sort.Slice(out, func(i, j int) bool { return out[i].Host < out[j].Host }) |
| 122 | return out, errors.Join(errs...) |
| 123 | } |
| 124 | func WriteStatus(ctx context.Context, s release.Store, acct, app string, st Status) (claimResult0 error) { |
| 125 | claimResult0 = release.WithAppMutation(ctx, s, acct, app, "schedule.WriteStatus", func(claimContext context.Context) error { |
| 126 | claimResult0 = claimRawWriteStatus(claimContext, s, acct, app, st) |
| 127 | return claimResult0 |
| 128 | }) |
| 129 | return |
| 130 | } |
| 131 | |
amadan