MIME-Version: 1.0
In-Reply-To: <5031a50c-aca7-4512-9478-95583606b5ff@Spark>
References: <5031a50c-aca7-4512-9478-95583606b5ff@Spark>
From: "David G. Johnston" <david.g.johnston@gmail.com>
Date: Sat, 22 Nov 2025 06:34:38 -0600
Message-ID: <CAKFQuwaWXkQ8Z5r+5n=H4JOHD4RkrxOJB6DNB0ha1qTP4u4aZQ@mail.gmail.com>
Subject: Re: FK from logged to unlogged table?
To: "dolan@directdemocracysolutions.com" <dolan@directdemocracysolutions.com>
Cc: "pgsql-general@lists.postgresql.org" <pgsql-general@lists.postgresql.org>
Content-Type: multipart/alternative; boundary="0000000000005421c706442e2736"
Archived-At: <https://www.postgresql.org/message-id/CAKFQuwaWXkQ8Z5r%2B5n%3DH4JOHD4RkrxOJB6DNB0ha1qTP4u4aZQ%40mail.gmail.com>
Precedence: bulk

--0000000000005421c706442e2736
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Friday, November 21, 2025, <dolan@directdemocracysolutions.com> wrote:

> Hello,
>
> I'm looking to improve bulk write performance on a table of about 23
> million rows by setting it unlogged. If lost, the table can be re-generat=
ed
> from archived raw data. The unlogged table would be referenced from a
> different, logged, table by a sparse but very important foreign key.
>
> If the unlogged table is lost, I can repair the foreign key data by
> re-uploading the raw data and following a different unique key. However,
> this would be annoying and I would rather not have to implement it if I c=
an
> protect the keyed records instead.
>
>    1. Is it allowed to key from a logged table to an unlogged table?
>    2. What is the system behavior if the unlogged table is lost?
>    3. Is there a clean way to protect only the subset of records that are
>    keyed? (Yes, I am considering periodic backups to an archive table, bu=
t
>    there would still be some repair needed that way).
>
>
Both 1 and 2 Feels like something you should take the couple of minutes to
try.  For 3, it=E2=80=99s an unlogged table - the entire thing.

David J.

--0000000000005421c706442e2736
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Friday, November 21, 2025,  &lt;<a href=3D"mailto:dolan@directdemocracys=
olutions.com">dolan@directdemocracysolutions.com</a>&gt; wrote:<br><blockqu=
ote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc s=
olid;padding-left:1ex">


<div>
<div name=3D"messageBodySection">
<div>Hello,</div>
<div>=C2=A0</div>
<div>I&#39;m looking to improve bulk write performance on a table of about =
23 million rows by setting it unlogged. If lost, the table can be re-genera=
ted from archived raw data. The unlogged table would be referenced from a d=
ifferent, logged, table by a sparse but very important foreign key.</div>
<div>=C2=A0</div>
<div>If the unlogged table is lost, I can repair the foreign key data by re=
-uploading the raw data and following a different unique key. However, this=
 would be annoying and I would rather not have to implement it if I can pro=
tect the keyed records instead.</div>
<ol type=3D"1">
<li>Is it allowed to key from a logged table to an unlogged table?</li>
<li>What is the system behavior if the unlogged table is lost?</li>
<li>Is there a clean way to protect only the subset of records that are key=
ed? (Yes, I am considering periodic backups to an archive table, but there =
would still be some repair needed that way).</li></ol></div></div></blockqu=
ote><div><br></div><div>Both 1 and 2 Feels like something you should take t=
he couple of minutes to try.=C2=A0 For 3, it=E2=80=99s an unlogged table - =
the entire thing.</div><div><br></div><div>David J.</div><div><br></div>

--0000000000005421c706442e2736--