Re: Historic snapshot doesn't track txns committed in BUILDING_SNAPSHOT state

Hi,

On Mon, Aug 12, 2024 at 04:34:25PM +0800, cca5507 wrote:
> Hi,
>
>
> 4, 5 ===
>
>
> &gt; if (SnapBuildCurrentState(builder) < SNAPBUILD_BUILDING_SNAPSHOT ||
> &gt;&nbsp; &nbsp; &nbsp;(SnapBuildCurrentState(builder) == SNAPBUILD_BUILDING_SNAPSHOT &amp;&amp; info != XLOG_HEAP_INPLACE) ||
> &gt;&nbsp; &nbsp; &nbsp;ctx-&gt;fast_forward)
> &gt;&nbsp; &nbsp; &nbsp;return;
>
>
>
> I think during fast forward, we also need handle the xlog that marks a transaction
> as&nbsp;catalog modifying, or the snapshot might lose some transactions?

I think it's fine to skip during fast forward as we are not generating logical
changes. It's done that way in master, in your proposal and in my "if" proposals.
Note that my proposals related to the if conditions are for heap2_decode and
heap_decode (not xact_decode).

> &gt; That way we'd still rely on what's being done in the XLOG_HEAP_INPLACE case
>
>
> + if (SnapBuildCurrentState(builder) &gt;= SNAPBUILD_BUILDING_SNAPSHOT)
> + {
> + /* Currently only XLOG_HEAP_INPLACE means a catalog modifying */
> + if (info == XLOG_HEAP_INPLACE &amp;&amp; TransactionIdIsValid(xid))
> + ReorderBufferXidSetCatalogChanges(ctx-&gt;reorder, xid, buf-&gt;origptr);
> + }
>
>
>
> We only call&nbsp;ReorderBufferXidSetCatalogChanges() for the xlog that marks a transaction as&nbsp;catalog
> modifying, and we don't care about the other steps being done in the xlog, so I think the current
> approach is ok.

Yeah, I think your proposal does not do anything wrong. I just prefer to put
everything in a single if condition (as per my proposal) so that we can jump
directly in the appropriate case. I think that way the code is easier to maintain
instead of having to deal with the same info values in distinct places.

Regards,

--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com