Re: Unlogged bloom indexes break after crash recovery

From: Heikki Linnakangas <hlinnaka(at)iki(dot)fi>
To: Narek Galstyan <narek(dot)galstyan(at)enterprisedb(dot)com>, pgsql-hackers(at)postgresql(dot)org
Subject: Re: Unlogged bloom indexes break after crash recovery
Date: 2026-11-05 14:41:31
Message-ID: fa94c943-c384-417c-afef-4323ebead763@iki.fi
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 04/11/2026 23:30, Narek Galstyan wrote:
> Extensions in the wild (e.g. pgvector here <https://github.com/
> pgvector/pgvector/blob/b0665dca96f0a186d438adbd1c35945d71da5490/src/
> ivfbuild.c#L1047-L1049> and a few private ones I know)
> assume GenericXLog does not log the init fork and they manually log
> the init fork with logic similar to the linked pgvector ivfflat.
> This means that for pgvector ivfflat, the potentially substantial
> ivfflat init fork (~8MB for lists=1000, dim = 1536) will be WAL-
> logged twice. This is pretty harmless. Still, probably worth
> addressing in pgvector. Notably, pgvector hnsw does not have the
> double logging problem because it does not use GenericXLog for the
> initial build.

Hmm, ok, that gives me pause. Perhaps best to do this in master only,
and use your fix for stable branches.

- Heikki

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Ayoub Kazar 2026-11-05 14:52:23 Re: Fix exclude generated columns in COPY TO with RLS
Previous Message Fujii Masao 2026-11-05 14:37:16 Re: [Bug Report + Patch] File descriptor leak when io_method=io_uring