Hi, On Wed, Oct 22, 2025 at 02:51:31PM +0900, Michael Paquier wrote: > > A first part here is something that Bertrand Drouvot has been working > on: being able to rebuild the table stats during WAL replay requires > us to move the stats to be relfilenode-based, so as the startup could > know where to recreate some of the numbers lost during crash recovery > from WAL, giving autovacuum a safety net.
Yeah, I resumed working on it recently and should be able to share a patch next week (based on what we discussed in [1]). Next week patch will be the first step i.e replace (i.e get rid) PGSTAT_KIND_RELATION by a brand new PGSTAT_KIND_RELFILENODE and move all the existing stats that are currently under the PGSTAT_KIND_RELATION to this new PGSTAT_KIND_RELFILENODE. It will do this by keeping the pg_stat_all_tables|indexes and pg_statio_all_tables|indexes on top of the PGSTAT_KIND_RELFILENODE and ensure that a relation rewrite keeps those stats. Next step will be that, once done, we could work from there to add new stats (add writes counters (heap_blks_written and friends) and ensure that some counters (n_dead_tup and friends) are replicated). [1]: https://www.postgresql.org/message-id/aN07p71KNFR2HdaD%40ip-10-97-1-34.eu-west-3.compute.internal Regards, -- Bertrand Drouvot PostgreSQL Contributors Team RDS Open Source Databases Amazon Web Services: https://aws.amazon.com
