[Bizgres-general] [ENG] Re: Statement Queuing take II - Resource Scheduling

[Mark Kirkwood]

Mark Kirkwood wrote:
>
>
> It is obvious in hindsight - changing the granularity from queue to 
> (queue, statement id) means that the wakeup routine also needs to 
> change, it must look at waiters on *all* locks for the same queue now... 
> I'm investigating this, but I'm unsure if can be made to work sensibly - 
> if not I'll need to leave proclocks alone and devise another data 
> structure to track (process, queue, statement).
>
>   
To fix this I think I'd need another data structure to track which 
statement ids either have or are waiting for locks on the same queue.

Given I need to add a new structure whatever design is used, I will go 
back to 1 lock for 1 queue and represent the relationship of 1 process 
having many statements in another structure - this is the more natural 
of the two approaches I think.

The good news is that the most of yesterdays work on locking by 
statement is not wasted, as I need to track statement ids in the new 
structure!

Cheers

Mark