Re: Query with AFTER UPDATE trigger not visible in binary logs

[Peter Brawley]

OK, I don't have an answer for #1, I'd say answer #2 is no unless some changes are failing to reach the slave(s), answer #3 is that there's a transaction even when you've not written one, ie row locks and gap locks need to be applied for your multi-row delete too, so the deadlock doesn't surprise me. The bigger the transaction, the bigger the risk of deadlock, so it appears you need to slim down the transaction or add deadlock-handling logic.