Re: [PATCH] Refactoring of LWLock tranches

On Thu, Nov 19, 2015 at 4:26 PM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> On 2015-11-19 14:58:05 -0500, Robert Haas wrote:
>> On Thu, Nov 19, 2015 at 11:30 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
>> > It's really not particularly convenient to allocate tranches right
>> > now. You have to store at least the identifier in shared memory and
>> > then redo the registration in each process. Otherwise some processes
>> > can't identify them. Which of rather inconvenient of you want to
>> > register some at runtime
>>
>> Sure, that's why we're proposing to use an enum or a list of #defines
>> for that. I don't see a need to do any more than that.
>
> That works fine for builtin stuff, but not at all for extensions doing
> it.
>
> If you do register locks at runtime, instead of shared_preload_library -
> something you surely agree makes some things easier by not requiring a
> restart - you really don't have any way to force the registration to
> happen in each backend.

I'm not exactly clear what you are proposing. Pick some relatively
small cap on the number of tranches and put an array of them in shared
memory? Blech.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company