On Mon, Oct 23, 2023 at 8:43 PM Andres Freund wrote:
> The source of the emode=13=DEBUG2 is that that's hardcoded in
> WaitForWALToBecomeAvailable(). I guess the error ought to come from
> XLogPageRead(), but all that happens is this:
>
> case XLREAD_FAIL:
>
Hi,
On 2023-10-23 16:21:45 -0700, Andres Freund wrote:
> 1) For some reason I haven't yet debugged, the ReadRecord(PANIC) in
>PerformWalRecovery() doesn't PANIC, but instead just returns NULL
>
>We *do* output a DEBUG message, but well, that's insufficient.
The debug is from this backtrac
Hi,
I investigated a crashing postgres instance. Turns out the original issue was
operator error. But in the process I found a few postgres issues. The scenario
is basically that redo LSN and checkpoint LSN are in seperate segments, and
that for whatever reason, the file containing the redo LSN is
