Replication lag when no primary key on InnoDB

[Stéphane Brisson]

Hi,

Using:
mysql Ver 14.14 Distrib 5.1.49, for unknown-linux-gnu (x86_64) using readline 5.1


Deleting rows from a master is causing replication lag when there is no PRIMARY KEY on InnoDB table. When I delete a small subset ( < 500) , everything is fine (slave can handle it).

But when it's a massive delete, > 5000 rows, Seconds_Behind_Master keeps increasing... and slave thread seems to get stuck on 'Reading event from the relay log'.

I'm asking why it's taking so long. It's look like the thread is doing a full table scan on each row deletion because there's no primary key. Am I right ?

Was fine when it was in statement-based binary logging in 5.0.

Thanks

Stephan