Horton

Thread

Commits

  1. Document relaxed HOT for summarizing indexes

  1. HOT readme missing documentation on summarizing index handling

    Matthias van de Meent <boekewurm+postgres@gmail.com> — 2023-07-06T11:40:31Z

    Hi,

With PG16's 19d8e230, we got rid of BRIN's blocking of HOT updates,
but I just realized that we failed to update the README.HOT document
with this new exception for summarizing indexes.

Attached a patch that updates that document, detailing the related rationale.

I'm not sure if such internal documentation is relevant for
backpatching, but I also don't think it woudl hurt to have this
included in the REL_16_STABLE branch.

Kind regards,

Matthias van de Meent
Neon (https://neon.tech/)

  2. Re: HOT readme missing documentation on summarizing index handling

    Tomas Vondra <tomas.vondra@enterprisedb.com> — 2023-07-06T22:14:42Z

    Yeah, README.HOT should have been updated, and I see no reason not to
backpatch this to v16. Barring objections, I'll do that tomorrow.

I have two suggesting regarding the README.HOT changes:

1) I'm not entirely sure it's very clear what "referential integrity of
indexes across tuple updates" actually means. I'm afraid "referential
integrity" may lead readers to think about foreign keys. Maybe it'd be
better to explain this is about having index pointers to the new tuple
version, etc.

2) Wouldn't it be good to make it a bit more explicit we now have three
"levels" of HOT:

  (a) no indexes need update
  (b) update only summarizing indexes
  (c) update all indexes

The original text was really about on/off, and I'm not quite sure the
part about "exception" makes this very clear.


regards

-- 
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

  3. Re: HOT readme missing documentation on summarizing index handling

    Matthias van de Meent <boekewurm+postgres@gmail.com> — 2023-07-07T16:34:09Z

    On Fri, 7 Jul 2023 at 00:14, Tomas Vondra <tomas.vondra@enterprisedb.com> wrote:
>
> Yeah, README.HOT should have been updated, and I see no reason not to
> backpatch this to v16. Barring objections, I'll do that tomorrow.
>
> I have two suggesting regarding the README.HOT changes:
>
> 1) I'm not entirely sure it's very clear what "referential integrity of
> indexes across tuple updates" actually means. I'm afraid "referential
> integrity" may lead readers to think about foreign keys. Maybe it'd be
> better to explain this is about having index pointers to the new tuple
> version, etc.
>
> 2) Wouldn't it be good to make it a bit more explicit we now have three
> "levels" of HOT:
>
>   (a) no indexes need update
>   (b) update only summarizing indexes
>   (c) update all indexes
>
> The original text was really about on/off, and I'm not quite sure the
> part about "exception" makes this very clear.

Agreed on both points. Attached an updated version which incorporates
your points.

Kind regards,

Matthias van de Meent
Neon (https://neon.tech)

  4. Re: HOT readme missing documentation on summarizing index handling

    Tomas Vondra <tomas.vondra@enterprisedb.com> — 2023-07-07T17:06:35Z

    On 7/7/23 18:34, Matthias van de Meent wrote:
> On Fri, 7 Jul 2023 at 00:14, Tomas Vondra <tomas.vondra@enterprisedb.com> wrote:
>>
>> Yeah, README.HOT should have been updated, and I see no reason not to
>> backpatch this to v16. Barring objections, I'll do that tomorrow.
>>
>> I have two suggesting regarding the README.HOT changes:
>>
>> 1) I'm not entirely sure it's very clear what "referential integrity of
>> indexes across tuple updates" actually means. I'm afraid "referential
>> integrity" may lead readers to think about foreign keys. Maybe it'd be
>> better to explain this is about having index pointers to the new tuple
>> version, etc.
>>
>> 2) Wouldn't it be good to make it a bit more explicit we now have three
>> "levels" of HOT:
>>
>>   (a) no indexes need update
>>   (b) update only summarizing indexes
>>   (c) update all indexes
>>
>> The original text was really about on/off, and I'm not quite sure the
>> part about "exception" makes this very clear.
> 
> Agreed on both points. Attached an updated version which incorporates
> your points.
> 

Thanks, pushed after correcting a couple typos.


regards

-- 
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

  5. Re: HOT readme missing documentation on summarizing index handling

    Matthias van de Meent <boekewurm+postgres@gmail.com> — 2023-07-07T17:12:31Z

    On Fri, 7 Jul 2023 at 19:06, Tomas Vondra <tomas.vondra@enterprisedb.com> wrote:
>
> On 7/7/23 18:34, Matthias van de Meent wrote:
> > On Fri, 7 Jul 2023 at 00:14, Tomas Vondra <tomas.vondra@enterprisedb.com> wrote:
> >> The original text was really about on/off, and I'm not quite sure the
> >> part about "exception" makes this very clear.
> >
> > Agreed on both points. Attached an updated version which incorporates
> > your points.
> >
>
> Thanks, pushed after correcting a couple typos.

Thanks!

Kind regards,

Matthias van de Meent
Neon (https://neon.tech)

  6. Re: HOT readme missing documentation on summarizing index handling

    Aleksander Alekseev <aleksander@timescale.com> — 2023-07-10T12:33:27Z

    Hi,

> > Thanks, pushed after correcting a couple typos.
>
> Thanks!

I noticed that ec99d6e9c87a introduced a slight typo:

s/if there is not room/if there is no room

-- 
Best regards,
Aleksander Alekseev