On Mon, 24 Nov 2025 at 23:09, Peter Geoghegan <pg@bowt.ie> wrote:
>
> On Mon, Nov 24, 2025 at 4:18 PM Hannu Krosing <hannuk@google.com> wrote:
> > When VACUUM decides which rows are safe to freeze or permanently
> > remove it currently ignores backends which have PROC_IN_VACUUM or
> > PROC_IN_LOGICAL_DECODING bits set.
> >
> > This patch adds PROC_IN_SAFE_IC to this set, so backends running
> > CREATE INDEX CONCURRENTLY or REINDEX CONCURRENTLY and where the index
> > is "simple" - i.e. not expression indexes or conditional indexes are
> > involved - these would be ignored too.
>
> Are you aware of commit d9d076222f5b? It was subsequently reverted by
> commit e28bb885 because it led to subtle data corruption. Indexes had
> wrong contents due to an unforeseen interaction with pruning.
Indeed, I don't think this is a correct change, given that these
visibility horizons are calculated in every backend, and are also used
for pruning.
On-access pruning is one case where this is used and which would break
- exactly the issue that caused d9d076222f5b to be reverted in
e28bb885.
Kind regards,
Matthias van de Meent
Databricks (https://www.databricks.com)
