On 2021-Nov-25, Tom Lane wrote:
> Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> > Oh, but also I think I should push a mitigation in case a production
> > system hits this problem: maybe reduce the message from FATAL to WARNING
> > if the registered LSN is at a page boundary.
>
> Uh, why? The fix should remove the problem, and if it doesn't, we're
> still looking at inconsistent WAL aren't we?
The problem is that the bug occurs while writing the WAL record. Fixed
servers won't produce such records, but if you run an unpatched server
and it happens to write one, without a mitigation you cannot get away
from FATAL during replay.
Since this bug exists in released minors, we should allow people to
upgrade to a newer version if they hit it.
--
Álvaro Herrera Valdivia, Chile — https://www.EnterpriseDB.com/
"Nadie está tan esclavizado como el que se cree libre no siéndolo" (Goethe)