| 1 | // The plans pass (plans-caps-packs spec §7, plans 2): learn whether plans are |
| 2 | // on for this deployment and, only if they are, read the limits record of |
| 3 | // every account this box serves into a last-known-good cache the activator |
| 4 | // asks. Rides the same passes as telemetryPass. |
| 5 | // |
| 6 | // Self-hosters are never interrupted: with no marker (no catalog on the |
| 7 | // console) the pass reads one object and applies nothing, a box with |
| 8 | // CARLOS_PLAN_LIMITS=off reads nothing at all, and an account with no record |
| 9 | // runs exactly as it always has. |
| 10 | package host |
| 11 | |
| 12 | import ( |
| 13 | "context" |
| 14 | "log" |
| 15 | "os" |
| 16 | "strings" |
| 17 | "sync" |
| 18 | "sync/atomic" |
| 19 | "time" |
| 20 | |
| 21 | "github.com/carlosframework/platform/internal/plans" |
| 22 | "github.com/carlosframework/platform/internal/registry" |
| 23 | "github.com/carlosframework/platform/internal/release" |
| 24 | ) |
| 25 | |
| 26 | // planLimitsSnapshot is immutable once stored; nil in planLimits means plans |
| 27 | // are off on this box. |
| 28 | type planLimitsSnapshot struct { |
| 29 | byAccount map[string]plans.Limits |
| 30 | } |
| 31 | |
| 32 | var planLimits atomic.Pointer[planLimitsSnapshot] |
| 33 | |
| 34 | // plansPassMu serialises whole passes: the full pass and the five-minute pass |
| 35 | // both run plansPass, and a stalled pass must not publish over a newer one |
| 36 | // (Astra plans-2 finding 2). Readers never take it. |
| 37 | var plansPassMu sync.Mutex |
| 38 | |
| 39 | // planLimitsFailures counts consecutive failing passes for log throttling. |
| 40 | var planLimitsFailures atomic.Int64 |
| 41 | |
| 42 | // planLimitsSwitchOn reads CARLOS_PLAN_LIMITS: off, false, 0 or no turn plans |
| 43 | // off on this box; anything else, including unset, leaves them to the marker. |
| 44 | func planLimitsSwitchOn(getenv func(string) string) bool { |
| 45 | switch strings.ToLower(strings.TrimSpace(getenv("CARLOS_PLAN_LIMITS"))) { |
| 46 | case "off", "false", "0", "no": |
| 47 | return false |
| 48 | } |
| 49 | return true |
| 50 | } |
| 51 | |
| 52 | // planLimitsFor is sqid's limits, or false when plans are off or the account |
| 53 | // has no record. |
| 54 | func planLimitsFor(sqid string) (plans.Limits, bool) { |
| 55 | snap := planLimits.Load() |
| 56 | if snap == nil { |
| 57 | return plans.Limits{}, false |
| 58 | } |
| 59 | l, ok := snap.byAccount[sqid] |
| 60 | return l, ok |
| 61 | } |
| 62 | |
| 63 | func plansPass(ctx context.Context, s release.Store, reg *registry.Registry, getenv func(string) string) { |
| 64 | plansPassMu.Lock() |
| 65 | defer plansPassMu.Unlock() |
| 66 | if getenv == nil { |
| 67 | getenv = os.Getenv |
| 68 | } |
| 69 | if !planLimitsSwitchOn(getenv) { |
| 70 | planLimits.Store(nil) |
| 71 | return |
| 72 | } |
| 73 | marker, _, err := plans.GetMarker(ctx, s) |
| 74 | if err != nil { |
| 75 | planLimitsFailed("marker: %v", err) |
| 76 | return // keep the last good cache |
| 77 | } |
| 78 | if marker == nil || !marker.Enabled { |
| 79 | planLimits.Store(nil) |
| 80 | planLimitsFailures.Store(0) |
| 81 | return |
| 82 | } |
| 83 | routes, err := reg.List() |
| 84 | if err != nil { |
| 85 | planLimitsFailed("routes: %v", err) |
| 86 | return |
| 87 | } |
| 88 | prev := planLimits.Load() |
| 89 | next := &planLimitsSnapshot{byAccount: map[string]plans.Limits{}} |
| 90 | seen := map[string]bool{} |
| 91 | failed := false |
| 92 | for _, rt := range routes { |
| 93 | if rt.Kind == "status" { |
| 94 | continue |
| 95 | } |
| 96 | sqid := release.RouteAccount(rt) |
| 97 | if sqid == "" || seen[sqid] { |
| 98 | continue |
| 99 | } |
| 100 | seen[sqid] = true |
| 101 | rec, _, err := plans.GetLimits(ctx, s, sqid) |
| 102 | switch { |
| 103 | case err != nil: |
| 104 | failed = true |
| 105 | planLimitsFailed("%s: %v", sqid, err) |
| 106 | if prev != nil { |
| 107 | if l, ok := prev.byAccount[sqid]; ok { |
| 108 | next.byAccount[sqid] = l |
| 109 | } |
| 110 | } |
| 111 | case rec != nil: |
| 112 | next.byAccount[sqid] = rec.Limits |
| 113 | } |
| 114 | } |
| 115 | planLimits.Store(next) |
| 116 | if !failed { |
| 117 | planLimitsFailures.Store(0) |
| 118 | } |
| 119 | } |
| 120 | |
| 121 | // planLimitsFailed logs the first failure of a run and every tenth after it. |
| 122 | func planLimitsFailed(format string, args ...any) { |
| 123 | if n := planLimitsFailures.Add(1); n == 1 || n%10 == 0 { |
| 124 | log.Printf("agent: plans pass: "+format+" (keeping the last good limits)", args...) |
| 125 | } |
| 126 | } |
| 127 | |
| 128 | // planIdleFor is the activator's plans idle hook: the route's account's idle |
| 129 | // window from its limits, or 0 (no opinion) when plans are off or the account |
| 130 | // has no record. |
| 131 | func planIdleFor(rt registry.Route) time.Duration { |
| 132 | l, ok := planLimitsFor(release.RouteAccount(rt)) |
| 133 | if !ok || l.IdleSeconds <= 0 { |
| 134 | return 0 |
| 135 | } |
| 136 | return time.Duration(l.IdleSeconds) * time.Second |
| 137 | } |
| 138 | |
amadan