Обсуждение: pg_walfile_name uses XLByteToPrevSeg
Hi All, pg_walfile_name() returns the WAL file name corresponding to the given WAL location. Per https://www.postgresql.org/docs/14/functions-admin.html --- pg_walfile_name ( lsn pg_lsn ) → text Converts a write-ahead log location to the name of the WAL file holding that location. --- The function uses XLByteToPrevSeg() which gives the name of previous WAL file if the location falls on the boundary of WAL segment. I find it misleading since the given LSN will fall into the segment provided by XLByteToSeg() and not XLBytePrevSeg(). And it gives some surprising results as well --- #select pg_walfile_name('0/0'::pg_lsn); pg_walfile_name -------------------------- 00000001FFFFFFFF000000FF (1 row) ---- Comment in the code says --- /* * Compute an xlog file name given a WAL location, * such as is returned by pg_stop_backup() or pg_switch_wal(). */ Datum pg_walfile_name(PG_FUNCTION_ARGS) --- XLByteToPrevSeg() may be inline with the comment but I don't think that's what is conveyed by the documentation at least. -- Best Wishes, Ashutosh
On Fri, Feb 4, 2022 at 9:05 AM Ashutosh Bapat
<ashutosh.bapat.oss@gmail.com> wrote:
> And it gives some surprising results as well
> ---
> #select pg_walfile_name('0/0'::pg_lsn);
> pg_walfile_name
> --------------------------
> 00000001FFFFFFFF000000FF
> (1 row)
> ----
Yeah, that seems wrong.
--
Robert Haas
EDB: http://www.enterprisedb.com
On Fri, Feb 04, 2022 at 09:17:54AM -0500, Robert Haas wrote:
> On Fri, Feb 4, 2022 at 9:05 AM Ashutosh Bapat
> <ashutosh.bapat.oss@gmail.com> wrote:
>> And it gives some surprising results as well
>> ---
>> #select pg_walfile_name('0/0'::pg_lsn);
>> pg_walfile_name
>> --------------------------
>> 00000001FFFFFFFF000000FF
>> (1 row)
>> ----
>
> Yeah, that seems wrong.
It looks like it's been this way for a while (704ddaa).
pg_walfile_name_offset() has the following comment:
* Note that a location exactly at a segment boundary is taken to be in
* the previous segment. This is usually the right thing, since the
* expected usage is to determine which xlog file(s) are ready to archive.
I see a couple of discussions about this as well [0] [1].
[0] https://www.postgresql.org/message-id/flat/1154384790.3226.21.camel%40localhost.localdomain
[1] https://www.postgresql.org/message-id/flat/15952.1154827205%40sss.pgh.pa.us
--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com
At Fri, 4 Feb 2022 14:50:57 -0800, Nathan Bossart <nathandbossart@gmail.com> wrote in
> On Fri, Feb 04, 2022 at 09:17:54AM -0500, Robert Haas wrote:
> > On Fri, Feb 4, 2022 at 9:05 AM Ashutosh Bapat
> > <ashutosh.bapat.oss@gmail.com> wrote:
> >> And it gives some surprising results as well
> >> ---
> >> #select pg_walfile_name('0/0'::pg_lsn);
> >> pg_walfile_name
> >> --------------------------
> >> 00000001FFFFFFFF000000FF
> >> (1 row)
> >> ----
> >
> > Yeah, that seems wrong.
>
> It looks like it's been this way for a while (704ddaa).
> pg_walfile_name_offset() has the following comment:
>
> * Note that a location exactly at a segment boundary is taken to be in
> * the previous segment. This is usually the right thing, since the
> * expected usage is to determine which xlog file(s) are ready to archive.
>
> I see a couple of discussions about this as well [0] [1].
>
> [0] https://www.postgresql.org/message-id/flat/1154384790.3226.21.camel%40localhost.localdomain
> [1] https://www.postgresql.org/message-id/flat/15952.1154827205%40sss.pgh.pa.us
Yes, its the deliberate choice of design, or a kind of
questionable-but-unoverturnable decision. I think there are many
external tools conscious of this behavior.
It is also described in the documentation.
https://www.postgresql.org/docs/current/functions-admin.html
> When the given write-ahead log location is exactly at a write-ahead
> log file boundary, both these functions return the name of the
> preceding write-ahead log file. This is usually the desired behavior
> for managing write-ahead log archiving behavior, since the preceding
> file is the last one that currently needs to be archived.
regards.
--
Kyotaro Horiguchi
NTT Open Source Software Center
At Mon, 07 Feb 2022 13:21:53 +0900 (JST), Kyotaro Horiguchi <horikyota.ntt@gmail.com> wrote in
> At Fri, 4 Feb 2022 14:50:57 -0800, Nathan Bossart <nathandbossart@gmail.com> wrote in
> > On Fri, Feb 04, 2022 at 09:17:54AM -0500, Robert Haas wrote:
> > > On Fri, Feb 4, 2022 at 9:05 AM Ashutosh Bapat
> > > <ashutosh.bapat.oss@gmail.com> wrote:
> > >> And it gives some surprising results as well
> > >> ---
> > >> #select pg_walfile_name('0/0'::pg_lsn);
> > >> pg_walfile_name
> > >> --------------------------
> > >> 00000001FFFFFFFF000000FF
> > >> (1 row)
> > >> ----
> > >
> > > Yeah, that seems wrong.
> >
> > It looks like it's been this way for a while (704ddaa).
> > pg_walfile_name_offset() has the following comment:
> >
> > * Note that a location exactly at a segment boundary is taken to be in
> > * the previous segment. This is usually the right thing, since the
> > * expected usage is to determine which xlog file(s) are ready to archive.
> >
> > I see a couple of discussions about this as well [0] [1].
> >
> > [0] https://www.postgresql.org/message-id/flat/1154384790.3226.21.camel%40localhost.localdomain
> > [1] https://www.postgresql.org/message-id/flat/15952.1154827205%40sss.pgh.pa.us
>
> Yes, its the deliberate choice of design, or a kind of
> questionable-but-unoverturnable decision. I think there are many
> external tools conscious of this behavior.
>
> It is also described in the documentation.
>
> https://www.postgresql.org/docs/current/functions-admin.html
> > When the given write-ahead log location is exactly at a write-ahead
> > log file boundary, both these functions return the name of the
> > preceding write-ahead log file. This is usually the desired behavior
> > for managing write-ahead log archiving behavior, since the preceding
> > file is the last one that currently needs to be archived.
I forgot to mentino, but I don't think we need to handle the
wrap-around case of the function.
regards.
--
Kyotaro Horiguchi
NTT Open Source Software Center