Date: Thu, 27 Jan 2022 16:24:34 +0900 (JST)
Message-Id: <20220127.162434.876092147237035105.horikyota.ntt@gmail.com>
To: michael@paquier.xyz
Cc: sulamul@gmail.com, rjuju123@gmail.com, andres@anarazel.de,
 pgsql-hackers@lists.postgresql.org
Subject: Re: TAP test to cover "EndOfLogTLI != replayTLI" case
From: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
In-Reply-To: <YfI8SS6cccRFWU2w@paquier.xyz>
References: <20220118.140123.342853390774842194.horikyota.ntt@gmail.com>
	<CAAJ_b94hsL4_aboiyPipfAEMzaW=OVFSd5h7sqCc+R49f3tY1g@mail.gmail.com>
	<YfI8SS6cccRFWU2w@paquier.xyz>
User-Agent: Mew version 6.8 on Emacs 26.1
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Archived-At: <https://www.postgresql.org/message-id/20220127.162434.876092147237035105.horikyota.ntt%40gmail.com>
Precedence: bulk

At Thu, 27 Jan 2022 15:31:37 +0900, Michael Paquier <michael@paquier.xyz> wrote in 
> On Thu, Jan 20, 2022 at 12:13:08PM +0530, Amul Sul wrote:
> > I found the cause for the test failing on window -- is due to the
> > custom archive command setting which wasn't setting the correct window
> > archive directory path.
> 
> After reading this patch and this thread, I have noticed that you are
> testing the same thing as Heikki here, patch 0001:
> https://www.postgresql.org/message-id/52bc9ccd-8591-431b-0086-15d9acf25a3f@iki.fi
> 
> The patch sent on the other thread has a better description and shape,
> so perhaps we'd better drop what is posted here in favor of the other
> version.  Thoughts?

pg_switch_wal() is more preferable than filling-in a large number of
records as the means to advance segments because it is stable against
changes of wal_segment_size.

As you said, using has_restoring on initializing node_ptr is simpler
than explicitly setting archive_dir from that of another node.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center