> Using pg_control to get checkpoint position speed up things but > to handle possible pg_control corruption we obviously should > implement reading existent log segments (from the last one - > newest - to oldest) to get last checkpoint. I think this would be *very* important :-) Andreas
