Re: PENDING_LIST_CLEANUP_SIZE - maximum size of GIN pending list Re: HEAD seems to generate larger WAL regarding GIN index

(2014/10/30 21:30), Fujii Masao wrote:
> On Thu, Oct 30, 2014 at 7:30 PM, Etsuro Fujita
> <fujita.etsuro@lab.ntt.co.jp> wrote:

>> Here are my review comments.
>>
>> * The patch applies cleanly and make and make check run successfully.  I
>> think that the patch is mostly good.
>
> Thanks! Attached is the updated version of the patch.

Thank you for updating the patch!

>> * In src/backend/utils/misc/guc.c:
>> +       {
>> +               {"pending_list_cleanup_size", PGC_USERSET,
>> CLIENT_CONN_STATEMENT,
>> +                       gettext_noop("Sets the maximum size of the pending
>> list for GIN index."),
>> +                        NULL,
>> +                       GUC_UNIT_KB
>> +               },
>> +               &pending_list_cleanup_size,
>> +               4096, 0, MAX_KILOBYTES,
>> +               NULL, NULL, NULL
>> +       },
>>
>> ISTM it'd be better to use RESOURCES_MEM, not CLIENT_CONN_STATEMENT. No?
>
> Yes if the pending list always exists in the memory. But not, IIUC. Thought?

Exactly.  But I think we can expect that in many cases, since I think 
that the users would often set the GUC to a small value to the extent 
that most of the pending list pages would be cached by shared buffer, to 
maintain *search* performance.

I'd like to hear the opinions of others about the category for the GUC.

>> Also why not set min to 64, not to 0, in accoradance with that of work_mem?
>
> I'm OK to use 64. But I just chose 0 because I could not think of any reasonable
> reason why 64k is suitable as the minimum size of the pending list.
> IOW, I have no idea about whether it's reasonable to use the min value of
> work_mem as the min size of the pending list.

IIUC, I think that min = 0 disables fast update, so ISTM that it'd be 
appropriate to set min to some positive value.  And ISTM that the idea 
of using the min value of work_mem is not so bad.

>> * In doc/src/sgml/ref/create_index.sgml:
>> +     <term><literal>PENDING_LIST_CLEANUP_SIZE</></term>
>>
>> IMHO, it seems to me better for this variable to be in lowercase in
>> accordance with the GUC version.
>
> Using lowercase only for pending_list_cleanup_size and uppercase for
> other options
> looks strange to me. What about using lowercase for all the storage options?

+1

> I changed the document in that way.

*** 356,361 **** CREATE [ UNIQUE ] INDEX [ CONCURRENTLY ] [ <replaceable 
class="parameter">name</
--- 356,372 ----      </listitem>     </varlistentry>     </variablelist>
+    <variablelist>
+    <varlistentry>
+     <term><literal>PENDING_LIST_CLEANUP_SIZE</></term>

The above is still in uppercse.

Thanks,

Best regards,
Etsuro Fujita