Comment 5 for bug 988036

How is your simulated binlog stream handling transactions/locking? If it's holding locks on huge ranges of rows, then pt-osc --retries may need to be very high. Or the inverse might be happening: since pt-osc 2.1 adjusts its own chunk size, maybe it's using a very large chunk size and it can't get locks for all those rows. Along with increasing --retries, try lowering --chunk-time to keep the tool from selecting too many rows.