On Thu, Nov 20, 2014 at 12:12 PM, Jeff Davis <pgsql@j-davis.com> wrote:
> On Thu, 2014-11-20 at 11:22 -0500, Robert Haas wrote:
>> 2. Propagate pre-existing locks from the user backend to all the workers.
>>
>> I initially proposed #1, but now I think #2 solves more of the
>> problems for less code.
>
> OK. The primary concern with that is unintended consequences. But it's
> reasonable for you to ask for something more concrete. I will think on
> this more.
>
> A few things I'm thinking about now:
>
> * What do you mean by "pre-existing"? Locks existing prior to what
> event? (I don't think that's exactly what you meant.)
> * What's the conceptual difference between granting locks that would
> otherwise conflict with another process in the group (which is what this
> proposal is about) and having exactly the same set of locks? Is there
> any?
> * Let's say you have processes A1 and A2 in one group, and B. A1 and B
> both have an AccessShare lock, and A2 tries to acquire an exclusive
> lock. B is waiting on A2. That's still a deadlock, right?
I think I discussed all of these issues on the other thread already.
Am I wrong?
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
