Re: We need to log aborted autovacuums

Simon Riggs <simon@2ndQuadrant.com> writes:
> I'm fairly confused by this thread.

> We *do* emit a message when we cancel an autovacuum task. We went to a
> lot of trouble to do that. The message is DEBUG2, and says
> "sending cancel to blocking autovacuum pid =".

That doesn't necessarily match one-to-one with actual cancellations,
nor does it cover the case Greg is on about at the moment of an AV
worker being blocked indefinitely because it can't get the table
lock in the first place.

It might be an adequate substitute, but on the whole I agree with
the idea that it'd be better to have autovacuum log when it actually
cancels an operation, not when someone tries to cancel one.
        regards, tom lane