Re: [Proposal] Fully WAL logged CREATE DATABASE - No Checkpoints - Mailing list pgsql-hackers

From Robert Haas
Subject Re: [Proposal] Fully WAL logged CREATE DATABASE - No Checkpoints
Date
Msg-id CA+TgmoamCmeHQONKPk4w1vO+nzAS-bytO9uDhW6TYvNc7U5+ag@mail.gmail.com
Whole thread Raw
In response to Re: [Proposal] Fully WAL logged CREATE DATABASE - No Checkpoints  (Dilip Kumar <dilipbalaut@gmail.com>)
Responses Re: [Proposal] Fully WAL logged CREATE DATABASE - No Checkpoints  (Dilip Kumar <dilipbalaut@gmail.com>)
List pgsql-hackers
On Wed, Mar 23, 2022 at 4:42 AM Dilip Kumar <dilipbalaut@gmail.com> wrote:
> Okay this is an interesting point.  So one option is that in case of
> failure while using the wal log strategy we do not remove the database
> directory, because an abort transaction will take care of removing the
> relation file.  But then in failure case we will leave the orphaned
> database directory with version file and the relmap file.  Another
> option is to do the redundant cleanup as we are doing now.  Any other
> options?

I think our overriding goal should be to get everything using one
mechanism. It doesn't look straightforward to get everything to go
through the PendingRelDelete mechanism, because as you say, it can't
handle non-relation files or directories. However, what if we opt out
of that mechanism? We could do that either by not using
RelationCreateStorage() in the first place and directly calling
smgrcreate(), or by using RelationPreserveStorage() afterwards to yank
the file back out of the list.

-- 
Robert Haas
EDB: http://www.enterprisedb.com



pgsql-hackers by date:

Previous
From: Andrew Dunstan
Date:
Subject: Re: cpluspluscheck vs ICU
Next
From: Justin Pryzby
Date:
Subject: Re: SQL/JSON: functions