Re: [HACKERS] Setting pd_lower in GIN metapage

On Mon, Jun 26, 2017 at 3:54 PM, Amit Langote
<Langote_Amit_f8@lab.ntt.co.jp> wrote:
> On 2017/06/26 10:54, Michael Paquier wrote:
>> On Fri, Jun 23, 2017 at 11:17 AM, Amit Langote
>> <Langote_Amit_f8@lab.ntt.co.jp> wrote:
>>> That was it, thanks for the pointer.
>>
>> GinInitMetabuffer() sets up pd_lower and pd_upper anyway using
>> PageInit so the check of PageIsVerified is guaranteed to work in any
>> case. Upgraded pages will still have their pd_lower set to the
>> previous values, and new pages will have the optimization. So this
>> patch is actually harmless for past pages, while newer ones are seen
>> as more compressible.
>
> Right.
>
>>> Attached updated patch, which I confirmed, passes wal_consistency_check = gin.
>>
>> I have spent some time looking at this patch, playing with pg_upgrade
>> to check the state of the page upgraded. And this looks good to me.
>
> Thanks for the review.
>
>> One thing that I noticed is that this optimization could as well
>> happen for spgist meta pages. What do others think?
>
> I agree.  As Sawada-san mentioned, brin metapage code can use a similar patch.
>
> So attached are three patches for gin, brin, and sp-gist respectively.
> Both brin and sp-gist cases didn't require any special consideration for
> passing wal_consistency_checking, as the patch didn't cause brin and
> sp-gist metapages to become invalid when recreated on standby (remember
> that patch 0001 needed to be updated for that).
>

Thank you for the patches! I checked additional patches for brin and
spgist. They look good to me.

Regards,

--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center