Re: more descriptive message for process termination due to max_slot_wal_keep_size

Hi,

On 9/7/22 4:20 AM, Kyotaro Horiguchi wrote:
> (I noticed I sent a wrong version..)
>
> At Tue, 6 Sep 2022 10:54:35 +0200, "Drouvot, Bertrand"<bdrouvot(at)amazon(dot)com> wrote in
>> Thanks for the new patch version!. I did not realized (sorry about
>> that) that we'd need to expose byte_size_pretty(). Now I wonder if we
> I didn't think we need the units larger than MB, but I used
> pretty_print to prevent small number from rounding to exactly zero.

Yeah makes sense.

Also, rounding to zero wouldn't occur with "just" displaying "oldestLSN
- restart_lsn" (as proposed upthread).

> On
> the other hand, in typical cases it is longer than 6 digits in bytes,
> which is a bit hard to read a glance.

Yeah right, but that's already the case in some part of the code, like
for example in arrayfuncs.c:

                ereport(ERROR,
                        (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
                         errmsg("array size exceeds the maximum allowed
(%d)",
                                (int) MaxAllocSize)));

>> LOG: terminating process 16034 to release replication slot "rep1" because its restart_lsn 0/3158000 exceeds the limit by 15368192 bytes
>> should not simply report the number of bytes (like I can see it is
>> done in many places). So something like:
> ..
>> + (errmsg("terminating process %d to release replication slot \"%s\"
>> because its restart_lsn %X/%X exceeds the limit by %lu bytes",
> ..
>> and then forget about exposing/using byte_size_pretty() (that would be
>> more consistent with the same kind of reporting in the existing code).
>>
>> What do you think?
> An alterntive would be rounding up to the whole MB, or a sub-MB.
>
>> ereport(LOG,
>> (errmsg("terminating process %d to release replication slot \"%s\" because its restart_lsn %X/%X exceeds the limit by %.1lf MB",
>> active_pid, NameStr(slotname),
>> LSN_FORMAT_ARGS(restart_lsn),
>> /* round-up at sub-MB */
>> ceil((double) (oldestLSN - restart_lsn) / 1024 / 102.4) / 10),

typo "/ 102.4" ?

>> LOG: terminating process 49539 to release replication slot "rep1" because its restart_lsn 0/3038000 exceeds the limit by 15.8 MB
> If the distance were 1 byte, it is shown as "0.1 MB".

Right and I'm -1 on it, I think we should stick to the "pretty" or the
"bytes only" approach (my preference being the bytes only one).

Regards,

--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services:https://aws.amazon.com