Re: pg_wal/RECOVERYHISTORY file remains after archive recovery

From: Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com>
To: Fujii Masao <masao(dot)fujii(at)gmail(dot)com>
Cc: Michael Paquier <michael(at)paquier(dot)xyz>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: pg_wal/RECOVERYHISTORY file remains after archive recovery
Date: 2019-09-27 13:00:16
Message-ID: CAD21AoAYQH+DJSTniSzaRbDMD-+2R__rU=Rqwz=WbK4qDos9XQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Fri, Sep 27, 2019 at 8:41 PM Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
>
> On Fri, Sep 27, 2019 at 7:16 PM Michael Paquier <michael(at)paquier(dot)xyz> wrote:
> >
> > On Fri, Sep 27, 2019 at 05:58:21PM +0900, Fujii Masao wrote:
> > > So you think that it's better to remove them just after writeTimeLineHistory()?
> > > Per the following Sawada-san's comment, I was thinking that idea is basically
> > > not good. And, RemoveNonParentXlogFiles() also removes garbage files from
> > > pg_wal. It's simpler if similar codes exist near. Thought?
> >
> > Sawada-san's argument of upthread is that it is not good to put
> > exitArchiveRecovery() after writeTimeLineHIstory(), which is what
> > cbc55da has done per the reasons mentioned in the commit log, and we
> > should not change that.
> >
> > My argument is we know that RECOVERYXLOG and RECOVERYHISTORY are not
> > needed anymore at this stage of recovery, hence we had better remove
> > them as soon as possible. I am not convinced that it is a good idea
> > to move the cleanup close to RemoveNonParentXlogFiles(). First, this
> > is an entirely different part of the logic where the startup process
> > has already switched to a new timeline. Second, we add more steps
> > between the moment the two files are not needed and the moment they
> > are removed, so any failure in-between would cause those files to
> > still be there (we cannot say either that we will not manipulate this
> > code later on) and we don't want those files to lie around. So,
> > mentioning that we do the cleanup just after writeTimeLineHIstory()
> > because we don't need them anymore is more consistent with what has
> > been done for ages for the end of archive recovery, something that
> > cbc55da unfortunately broke.
>
> Ok, I have no objection to remove them just after writeTimeLineHistory().
>

I abandoned once to move the removal code to between
writeTimeLineHistory() and timeline switching because of expanding the
window but since unlink itself will complete within a very short time
it would not be problamatic much.

Attached the updated patch that just moves the removal code.

Regards,

--
Masahiko Sawada

Attachment Content-Type Size
v3_remove_recovered_historyfile.patch text/x-patch 1.5 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Alvaro Herrera 2019-09-27 13:22:10 Re: Attempt to consolidate reading of XLOG page
Previous Message Bruce Momjian 2019-09-27 12:51:40 Re: pg_upgrade fails with non-standard ACL