| 1 | // refusalWatch turns D1's refusal records (release.Refusal, written by the |
| 2 | // root pass) into one alert each and a status row each. It runs on the |
| 3 | // agent's metrics tick. A record older than refusalAlertWindow when first |
| 4 | // seen is not alerted, so an agent restart does not re-send old news. |
| 5 | package host |
| 6 | |
| 7 | import ( |
| 8 | "context" |
| 9 | "log" |
| 10 | "sync/atomic" |
| 11 | "time" |
| 12 | |
| 13 | "github.com/carlosframework/platform/internal/release" |
| 14 | "github.com/carlosframework/platform/internal/status" |
| 15 | ) |
| 16 | |
| 17 | const refusalAlertWindow = 30 * time.Minute |
| 18 | |
| 19 | type refusalWatch struct{ seen map[string]time.Time } |
| 20 | |
| 21 | func (w *refusalWatch) tick(ctx context.Context, bakeDir string, sinks *alertSinks, now time.Time) []status.ReleaseRefusal { |
| 22 | if w.seen == nil { |
| 23 | w.seen = map[string]time.Time{} |
| 24 | } |
| 25 | recs, err := release.ListRefusals(bakeDir) |
| 26 | if err != nil { |
| 27 | log.Printf("agent: refusals: %v", err) |
| 28 | return nil |
| 29 | } |
| 30 | // A move's held-stage refusals (D3-held) live in their own namespace |
| 31 | // (final review finding 3) and are surfaced as they were when they |
| 32 | // shared adoption's: a row and an alert each, seen under their own key. |
| 33 | held, err := release.ListHeldRefusals(bakeDir) |
| 34 | if err != nil { |
| 35 | log.Printf("agent: held refusals: %v", err) |
| 36 | } |
| 37 | var rows []status.ReleaseRefusal |
| 38 | for i, r := range append(recs, held...) { |
| 39 | row := status.ReleaseRefusal{Host: r.Host, Version: r.Version, Reason: r.Reason, RefusedAt: r.RefusedAt, Kind: r.Kind} |
| 40 | if !r.Recovered { |
| 41 | row.RecoveryError = recoveryError(r) |
| 42 | } |
| 43 | rows = append(rows, row) |
| 44 | key := r.Host |
| 45 | if i >= len(recs) { |
| 46 | key = "held/" + r.Host |
| 47 | } |
| 48 | if last, ok := w.seen[key]; ok && !r.RefusedAt.After(last) { |
| 49 | continue |
| 50 | } |
| 51 | w.seen[key] = r.RefusedAt |
| 52 | if now.Sub(r.RefusedAt) <= refusalAlertWindow { |
| 53 | sinks.deliverRelease(ctx, r) |
| 54 | } |
| 55 | } |
| 56 | return rows |
| 57 | } |
| 58 | |
| 59 | // releaseRefusals is the metrics loop's latest answer, read by the status |
| 60 | // loop's probe. They are separate goroutines, so this is the seam between |
| 61 | // them (the same shape as processAlertCounters). Nil until the first tick. |
| 62 | var releaseRefusals atomic.Pointer[[]status.ReleaseRefusal] |
| 63 | |
| 64 | // releaseRefusalProbe feeds status.Probes.ReleaseRefusals. |
| 65 | func releaseRefusalProbe() []status.ReleaseRefusal { |
| 66 | if p := releaseRefusals.Load(); p != nil { |
| 67 | return append([]status.ReleaseRefusal(nil), (*p)...) |
| 68 | } |
| 69 | return nil |
| 70 | } |
| 71 | |