fix(master): refuse jobs for revoked agents, signal revoked jobs, agent start time wins (#7)

internal/master/store/agents.go

@@ -63,33 +63,39 @@ func (s *Store) AgentByTokenHash(ctx context.Context, hash string) (model.Agent,
 	return scanAgent(s.pool.QueryRow(ctx, `SELECT `+agentColumns+` FROM agents WHERE token_hash = $1`, hash))
 }
 
-// RevokeAgent marks the agent revoked and its Compose and plain-container
-// services removed and its queued and running jobs failed, in one transaction. Swarm rows are shared with other
-// managers and left alone. Its token stops working immediately; closing a
-// live connection is the hub's job.
-func (s *Store) RevokeAgent(ctx context.Context, id int64, at time.Time) error {
+// RevokeAgent marks the agent revoked, its Compose and plain-container
+// services removed and its queued and running jobs failed, in one
+// transaction, and returns the IDs of the jobs it failed. Swarm rows are
+// shared with other managers and left alone. Its token stops working
+// immediately; closing a live connection is the hub's job.
+func (s *Store) RevokeAgent(ctx context.Context, id int64, at time.Time) ([]int64, error) {
 	tx, err := s.pool.Begin(ctx)
 	if err != nil {
-		return fmt.Errorf("revoke agent: begin: %w", err)
+		return nil, fmt.Errorf("revoke agent: begin: %w", err)
 	}
 	defer func() { _ = tx.Rollback(ctx) }()
 
 	if _, err := tx.Exec(ctx,
 		`UPDATE agents SET revoked_at = $2 WHERE id = $1 AND revoked_at IS NULL`, id, at); err != nil {
-		return fmt.Errorf("revoke agent: %w", err)
+		return nil, fmt.Errorf("revoke agent: %w", err)
 	}
 	if _, err := tx.Exec(ctx,
 		`UPDATE services SET removed_at = $2
 		 WHERE agent_id = $1 AND kind IN ('compose', 'container') AND removed_at IS NULL`, id, at); err != nil {
-		return fmt.Errorf("revoke agent: remove services: %w", err)
+		return nil, fmt.Errorf("revoke agent: remove services: %w", err)
 	}
-	if _, err := tx.Exec(ctx, failActiveJobsSQL, id, "agent revoked", at); err != nil {
-		return fmt.Errorf("revoke agent: fail jobs: %w", err)
+	rows, err := tx.Query(ctx, failActiveJobsSQL, id, "agent revoked", at)
+	if err != nil {
+		return nil, fmt.Errorf("revoke agent: fail jobs: %w", err)
+	}
+	failed, err := pgx.CollectRows(rows, pgx.RowTo[int64])
+	if err != nil {
+		return nil, fmt.Errorf("revoke agent: fail jobs: %w", err)
 	}
 	if err := tx.Commit(ctx); err != nil {
-		return fmt.Errorf("revoke agent: commit: %w", err)
+		return nil, fmt.Errorf("revoke agent: commit: %w", err)
 	}
-	return nil
+	return failed, nil
 }
 
 // RecordHello stores what the agent reported about itself on connect.