Re: PostgreSQL logical decoder output plugin - unchanged toast data

From: Georgy Buranov <gburanov(at)gmail(dot)com>
To: andres(at)anarazel(dot)de
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: PostgreSQL logical decoder output plugin - unchanged toast data
Date: 2018-08-31 14:55:37
Message-ID: CAGMDm35y_zzwEpcLUs=YGQeXEuEtW9M0OsrAhooFtFmXQTNkVA@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Ok, thank you very much for your explanation,

maybe I need something else in my case.

As far as I understand, "On-disk toasted data for tuples from the WAL
are not guaranteed in any way to be retain", but still, the LATEST
value for the same cell should exist in postgres (in on-disk toast if
it is huge, or not).

If I cannot get access to the on-disk toasted data for tuple from the
WAL, can I have the access to the _latest_ value in this case
(hopefully I describe it correct)

> Yes, that's not possible in general. On-disk toasted data for tuples
> from the WAL are not guaranteed in any way to be retained. If that
> weren't the case database tables would bloat while logical replication
> is behind, and the sequential reads (i.e. fast) reads of logical
> decoding would turn into random IO.

Mit freundlichen Grüßen,
Georgy Buranov

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2018-08-31 15:05:57 Re: remove ancient pre-dlopen dynloader code
Previous Message Tom Lane 2018-08-31 14:55:06 Re: pg_verify_checksums and -fno-strict-aliasing