Hi,
On 2023-08-09 08:03:29 +0900, Michael Paquier wrote:
> On Tue, Aug 08, 2023 at 03:59:54PM -0700, Andres Freund wrote:
> > On 2023-08-08 08:54:10 +0900, Michael Paquier wrote:
> >> - WaitEventExtensionShmemInit() should gain a dshash_create(), to make
> >> sure that the shared table is around, and we are going to have a
> >> reference to it in WaitEventExtensionCounterData, saved from
> >> dshash_get_hash_table_handle().
> >
> > I'm not even sure it's worth using dshash here. Why don't we just create a
> > decently sized dynahash (say 128 enties) in shared memory? We overallocate
> > shared memory by enough that there's a lot of headroom for further entries, in
> > the rare cases they're needed.
>
> The question here would be how many slots the most popular extensions
> actually need, but that could always be sized up based on the
> feedback.
On a default initdb (i.e. 128MB s_b), after explicitly disabling huge pages,
we over-allocate shared memory by by 1922304 bytes, according to
pg_shmem_allocations. We allow that memory to be used for things like shared
hashtables that grow beyond their initial size. So even if the hash table's
static size is too small, there's lots of room to grow, even on small systems.
Just because it's somewhat interesting: With huge pages available and not
disabled, we over-allocate by 3364096 bytes.
Greetings,
Andres Freund