Horton

Re: reloption to prevent VACUUM from truncating empty pages at the end of relation

Tom Lane <tgl@sss.pgh.pa.us>

View thread
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
Cc: Robert Haas <robertmhaas@gmail.com>, Michael Paquier <michael@paquier.xyz>, Julien Rouhaud <rjuju123@gmail.com>, "Tsunakawa, Takayuki" <tsunakawa.takay@jp.fujitsu.com>, Laurenz Albe <laurenz.albe@cybertec.at>, "Jamison, Kirk" <k.jamison@jp.fujitsu.com>, Alvaro Herrera <alvherre@2ndquadrant.com>, Amit Kapila <amit.kapila16@gmail.com>, Fujii Masao <masao.fujii@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2019-03-01T19:29:13Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Add TRUNCATE parameter to VACUUM.

  2. Add vacuum_truncate reloption.

  3. Allow VACUUM to be run with index cleanup disabled.

Andres Freund <andres@anarazel.de> writes:
> On 2019-03-01 14:17:33 -0500, Tom Lane wrote:
>> I think we should reject the whole patch, tbh, and go do something
>> about the underlying problem instead.  Once we've made truncation
>> not require AEL, this will be nothing but a legacy wart that we'll
>> have a hard time getting rid of.

> IDK, it's really painful in the field, and I'm not quite seeing us
> getting rid of the AEL for v12.

Dunno, I was musing about it just yesterday, in
https://postgr.es/m/1261.1551392263@sss.pgh.pa.us

I'd sure rather spend time making that happen than this.  I'm also
not entirely convinced that we MUST do something about this in v12
rather than v13 --- we've been living with it ever since we had
in-core replication, why's it suddenly so critical?

> I think it's a wart, but one that works
> around a pretty important usability issue. And I think we should just
> remove the GUC without any sort of deprecation afterwards, if necessary
> we can add a note to the docs to that effect.  It's not like preventing
> truncation from happening is a very intrusive / dangerous thing to do.

Well, if we add a reloption then we can never ever get rid of it; at
best we could ignore it.  So from the perspective of how-fast-can-we-
deprecate-this, maybe a GUC is the better answer.  On the other hand,
I'm not sure I believe that many installations could afford to disable
truncation for every single table.

			regards, tom lane