feat(jobs/runners): job_reaper (lease timeout) + job_purge (retention)

internal/jobs/runners/job_reaper.go

@@ -0,0 +1,43 @@
+package runners
+
+import (
+	"context"
+	"log/slog"
+	"time"
+
+	"github.com/google/uuid"
+)
+
+// JobsReapRepo is the minimal jobs repository surface for the reaper.
+type JobsReapRepo interface {
+	ReapStuckRunningJobs(ctx context.Context, leasedBefore time.Time) ([]uuid.UUID, error)
+}
+
+// JobReaper resets jobs.running rows whose lease has exceeded the timeout
+// (worker probably crashed). Periodic; idempotent; no retry needed.
+type JobReaper struct {
+	repo         JobsReapRepo
+	leaseTimeout time.Duration
+	log          *slog.Logger
+}
+
+// NewJobReaper builds the reaper.
+func NewJobReaper(repo JobsReapRepo, leaseTimeout time.Duration, log *slog.Logger) *JobReaper {
+	if log == nil {
+		log = slog.Default()
+	}
+	return &JobReaper{repo: repo, leaseTimeout: leaseTimeout, log: log}
+}
+
+// Run reaps stuck rows once.
+func (r *JobReaper) Run(ctx context.Context) {
+	cutoff := time.Now().Add(-r.leaseTimeout)
+	ids, err := r.repo.ReapStuckRunningJobs(ctx, cutoff)
+	if err != nil {
+		r.log.Error("job_reaper.reap", "err", err.Error())
+		return
+	}
+	if len(ids) > 0 {
+		r.log.Warn("job_reaper.reaped", "count", len(ids), "ids", ids)
+	}
+}