Re: PATCH: standby crashed when replay block which truncated in standby but failed to truncate in master node

On Fri, Jan 17, 2020 at 1:47 PM Michael Paquier <michael(at)paquier(dot)xyz> wrote:
>
> On Thu, Jan 16, 2020 at 11:17:36PM +0900, Fujii Masao wrote:
> > OK, I updated the patch that way.
> > Attached is the updated version of the patch.
>
> Thanks. I have few tweaks to propose to the docs.
>
> + raise a PANIC-level error, aborting the recovery. Setting
> Instead of "PANIC-level error", I would just use "PANIC error", and

I have no strong opinion about this, but I used "PANIC-level error"
because the description for data_sync_retry has already used it.

> instead of "aborting the recovery" just "crashing the server".

PANIC implies server crash, so IMO "crashing the server" is
a bit redundant, and "aborting the recovery" is better because
"continue the recovery" is used later.

> + causes the system to ignore those WAL records
> WAL records are not ignored, but errors caused by incorrect page
> references in those WAL records are. The current phrasing sounds like
> the WAL records are not applied.

So, what about

---------------
causes the system to ignore invalid page references in WAL records
(but still report a warning), and continue the recovery.
---------------

> Another thing that I just recalled. Do you think that it would be
> better to mention that invalid page references can only be seen after
> reaching the consistent point during recovery? The information given
> looks enough, but I was just wondering if that's worth documenting or
> not.

ISTM that this is not the information that users should understand...

Regards,

--
Fujii Masao