Re: A failure in t/038_save_logical_slots_shutdown.pl

On Wed, Jan 10, 2024 at 6:37 PM vignesh C <vignesh21(at)gmail(dot)com> wrote:
>
> I got the log files from Bharath offline. Thanks Bharath for sharing
> the log files offline.
> The WAL record sequence is exactly the same in the failing test and
> tests which are passing.
> One observation in our case the confirmed flush lsn points exactly to
> shutdown checkpoint, but in the failing test the lsn pointed is
> invalid, pg_waldump says that address is invalid and skips about 24
> bytes and then sees a valid record
>
> Passing case confirm flush lsn(0/150D158) from my machine:
> pg_waldump 000000010000000000000001 -s 0/150D158
> rmgr: XLOG len (rec/tot): 114/ 114, tx: 0, lsn:
> 0/0150D158, prev 0/0150D120, desc: CHECKPOINT_SHUTDOWN redo 0/150D158;
> tli 1; prev tli 1; fpw true; xid 0:739; oid 16388; multi 1; offset 0;
> oldest xid 728 in DB 1; oldest multi 1 in DB 1; oldest/newest commit
> timestamp xid: 0/0; oldest running xid 0; shutdown
>
> Failing case confirm flush lsn( 0/1508000) from failing tests log file:
> pg_waldump 000000010000000000000001 -s 0/1508000
> pg_waldump: first record is after 0/1508000, at 0/1508018, skipping
> over 24 bytes
> rmgr: XLOG len (rec/tot): 114/ 114, tx: 0, lsn:
> 0/01508018, prev 0/01507FC8, desc: CHECKPOINT_SHUTDOWN redo 0/1508018;
> tli 1; prev tli 1; fpw true; xid 0:739; oid 16388; multi 1; offset 0;
> oldest xid 728 in DB 1; oldest multi 1 in DB 1; oldest/newest commit
> timestamp xid: 0/0; oldest running xid 0; shutdown
>
> I'm still not sure why in this case, it is not exactly pointing to a
> valid WAL record, it has to skip 24 bytes to find the valid checkpoint
> shutdown record.
>

Can we see the previous record (as pointed out by prev in the WAL
record) in both cases? Also, you can see few prior records in both
cases.

--
With Regards,
Amit Kapila.