Re: PANIC during crash recovery of a recently promoted standby

From: Michael Paquier <michael(at)paquier(dot)xyz>
To: Kyotaro HORIGUCHI <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp>
Cc: pavan(dot)deolasee(at)gmail(dot)com, alvherre(at)2ndquadrant(dot)com, pgsql-hackers(at)postgresql(dot)org
Subject: Re: PANIC during crash recovery of a recently promoted standby
Date: 2018-06-20 06:21:42
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

On Thu, Jun 07, 2018 at 07:58:29PM +0900, Kyotaro HORIGUCHI wrote:
> Invalid-page checking during crash recovery is hamful rather than
> useless. It is done by CheckRecoveryConsistency even in crash
> recovery against its expectation because there's a case where
> minRecoveryPoint is valid but InArchiveRecovery is false. The two
> variable there seems to be in contradiction with each other.
> The immediate cause of the contradition is that StartXLOG wrongly
> initializes local minRecoveryPoint from control file's value even
> under crash recovery. updateMinRecoveryPoint also should be
> turned off during crash recovery. The case of crash after last
> checkpoint end has been treated in UpdateMinRecoveryPoint but
> forgot to consider this case, where crash recovery has been
> started while control file is still holding valid
> minRecoveryPoint.

Hmm. This patch looks interesting and those issues need a very careful
lookup. This is one of those things which should be fixed as part of
the extra CF, so I am planning to look at it in details very soon,
perhaps by the end of this week...

I have by the way noticed that nothing was registered in the CF app:
I have added as well a bullet point in the open item page of v11 for
older bugs.

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Michael Paquier 2018-06-20 06:27:14 Re: pg_config.h.win32 missing a set of flags from added in v11 development
Previous Message Michael Paquier 2018-06-20 06:09:55 Re: Adding tests for inheritance trees with temporary tables