Re: XID formatting and SLRU refactorings (was: Add 64-bit XIDs into PostgreSQL 15)

From: Aleksander Alekseev <aleksander(at)timescale(dot)com>
To: Postgres hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Cc: Heikki Linnakangas <hlinnaka(at)iki(dot)fi>, Maxim Orlov <orlovmg(at)gmail(dot)com>, Jacob Champion <jchampion(at)timescale(dot)com>, Japin Li <japinli(at)hotmail(dot)com>, Andres Freund <andres(at)anarazel(dot)de>, Michael Paquier <michael(at)paquier(dot)xyz>, Pavel Borisov <pashkin(dot)elfe(at)gmail(dot)com>, Peter Eisentraut <peter(dot)eisentraut(at)enterprisedb(dot)com>, Alexander Korotkov <aekorotkov(at)gmail(dot)com>
Subject: Re: XID formatting and SLRU refactorings (was: Add 64-bit XIDs into PostgreSQL 15)
Date: 2023-07-05 13:45:49
Message-ID: CAJ7c6TPmzoPgj-4zdf+UqdV4GcoZag+=KTLpSJ3RHp5pYSxfZg@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,

> Reviewing this now. I think it's almost ready to be committed.
>
> There's another big effort going on to move SLRUs to the regular buffer
> cache (https://commitfest.postgresql.org/43/3514/) I wonder how moving
> to 64 bit page numbers will affect that. BlockNumber is still 32 bits,
> after all.

Somehow I didn't pay too much attention to this effort, thanks. I will
familiarize myself with the patch. Intuitively I don't think that the
patchse should block each other.

> This patch makes the filenames always 12 characters wide (for SLRUs that
> opt-in to the new naming). That's actually not enough for the full range
> that a 64 bit page number can represent. Should we make it 16 characters
> now, to avoid repeating the same mistake we made earlier? Or make it
> more configurable, on a per-SLRU basis. One reason I don't want to just
> make it 16 characters is that WAL segment filenames are also 16 hex
> characters, which could cause confusion.

Good catch. I propose the following solution:

```
SlruFileName(SlruCtl ctl, char *path, int64 segno)
{
if (ctl->long_segment_names)
- return snprintf(path, MAXPGPATH, "%s/%012llX", ctl->Dir,
+ /*
+ * We could use 16 characters here but the disadvantage would be that
+ * the SLRU segments will be hard to distinguish from WAL segments.
+ *
+ * For this reason we use 15 characters. It is enough but also means
+ * that in the future we can't decrease SLRU_PAGES_PER_SEGMENT easily.
+ */
+ return snprintf(path, MAXPGPATH, "%s/%015llX", ctl->Dir,
(long long) segno);
else
return snprintf(path, MAXPGPATH, "%s/%04X", (ctl)->Dir,
```

PFE the corrected patchset v58.

--
Best regards,
Aleksander Alekseev

Attachment Content-Type Size
v58-0003-Make-use-FullTransactionId-in-2PC-filenames.patch application/octet-stream 2.9 KB
v58-0001-Index-SLRUs-by-64-bit-integers-rather-than-by-32.patch application/octet-stream 55.2 KB
v58-0002-Use-larger-segment-file-names-for-pg_notify.patch application/octet-stream 13.0 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Ashutosh Bapat 2023-07-05 13:50:06 Re: pg_decode_message vs skip_empty_xacts and xact_wrote_changes
Previous Message Euler Taveira 2023-07-05 13:37:05 Re: logicalrep_message_type throws an error