Re: BUG #16894: PANIC: WAL contains references to invalid pages

On Wed, Mar 03, 2021 at 06:49:02AM -0500, David Steele wrote:
> OK, but shouldn't we have a full page write for this page after the backup
> starts, rather than the partial we seem to be seeing here? Is there any
> condition where the full page write would be skipped legitimately, or does
> it point to a problem?

That's how things work. If they don't work this way for physical
backups, we may have a problem. At replay, the full page will be
replayed if BKPIMAGE_APPLY is correctly set, as per
XLogReadBufferForRedoExtended(). And XLogRecordAssemble()
does the decision when building the record (just grep for
needs_backup).

> If Postgres is running correctly there is certainly no expectation for
> support of this unusual use case, but I do think that this possibly points
> to an issue in Postgres, which under normal circumstances would be very hard
> to detect.

Well, the report tells that this is an issue that happens on those
fake files full of zeros, but are you sure that you have the sizing
right? I still don't see any evidence of anything broken based on the
information gathered for full backups, FWIW.
--
Michael