public inbox for [email protected]
help / color / mirror / Atom feedFrom: Aleksey M Boltenkov <[email protected]>
To: William Sescu (Suva) <[email protected]>
To: Raj <[email protected]>
Cc: Laurenz Albe <[email protected]>
Cc: Pgsql-admin <[email protected]>
Subject: Re: AW: Size of /pgdata
Date: Thu, 18 Jun 2026 23:33:20 +0300
Message-ID: <[email protected]> (raw)
In-Reply-To: <GV0P278MB0244AD72B327E57199FDD07CE8E32@GV0P278MB0244.CHEP278.PROD.OUTLOOK.COM>
References: <CAJk5Atb1v7=x-wsS-tL7Lxs7yXQy8Ng-GhXG+zzZ1gVAx=Xzjw@mail.gmail.com>
<[email protected]>
<CAJk5AtburOY2ugaH-6+8OHgOBwyyaq1V3LLozCAf57c5jtmR+w@mail.gmail.com>
<[email protected]>
<CAJk5AtZfFBqLfKyz7rAqJOXT0ZvEXna19wMn==YCin-PLUGiXw@mail.gmail.com>
<[email protected]>
<CAJk5AtZQ-4Wg0vnYLO5BJJntMG1oBEmvediAYQKXj0OL6MsOww@mail.gmail.com>
<[email protected]>
<CAJk5AtarTco=DXJw8qDOe_CET5hUApb-aEJr5xL2pEAFgnsdNQ@mail.gmail.com>
<[email protected]>
<GV0P278MB0244DE5A2049D0A42E471445E8E42@GV0P278MB0244.CHEP278.PROD.OUTLOOK.COM>
<CAJk5AtatQQ9npW-nffLXA9L-ivmyiEyvOse0XicbYZ53_-YATg@mail.gmail.com>
<GV0P278MB0244AD72B327E57199FDD07CE8E32@GV0P278MB0244.CHEP278.PROD.OUTLOOK.COM>
On 6/18/26 07:59, William Sescu (Suva) wrote:
> By setting XFS allocsize, you switch off the dynamic behavior, and it
> becomes more predictable.
>
> Another fix might be a switch to a different filesystem. E.g. ext4
>
> While this XFS feature is a quite old one, I have seen it the first time
> kicking in quite heavily only on Red Hat 9, with PostgreSQL 18 and
>
> only on the standby. Some times double the size for some files.
>
> *Von:*Raj <[email protected]>
> *Gesendet:* Mittwoch, 17. Juni 2026 20:46
> *An:* Sescu William (SW0) <[email protected]>
> *Cc:* Laurenz Albe <[email protected]>; Pgsql-admin <pgsql-
> [email protected]>
> *Betreff:* Re: Size of /pgdata
>
> Yes, this is what I have been asking. I also mentioned/asked earlier
> whether it's Abt XFS file system.
>
> I checked bitmaps and under flags I could find 'unwritten preallocated
> extent'. --apparent-size says shows correct size.
>
> Is the fix - setting up allocsize?
>
> On Wed, 17 Jun 2026, 19:29 William Sescu (Suva), <[email protected]
> <mailto:[email protected]>> wrote:
>
> In case you are using XFS, this might be related to the "Speculative
> preallocation" feature.
> We saw the same behavior on Redhat 9, with PostgreSQL 18 and XFS.
>
> https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/7/
> html/storage_administration_guide/migrating-ext4-xfs <https://
> docs.redhat.com/en/documentation/red_hat_enterprise_linux/7/html/
> storage_administration_guide/migrating-ext4-xfs>
>
> Speculative preallocation
>
> XFS uses speculative preallocation to allocate blocks past EOF as
> files are written. This avoids file fragmentation due to concurrent
> streaming write workloads on NFS servers. By default, this
> preallocation increases with the size of the file and will be
> apparent in "du" output. If a file with speculative preallocation is
> not dirtied for five minutes the preallocation will be discarded. If
> the inode is cycled out of cache before that time, then the
> preallocation will be discarded when the inode is reclaimed.
> If premature ENOSPC problems are seen due to speculative
> preallocation, a fixed preallocation amount may be specified with
> the -o allocsize=amount mount option.
>
> See also "man xfs"
>
> allocsize=size
> Sets the buffered I/O end-of-file preallocation size when doing
> delayed allocation writeout. Valid values for this option are page
> size (typically 4KiB) through to
> 1GiB, inclusive, in power-of-2 increments.
>
> The default behavior is for dynamic end-of-file preallocation size,
> which uses a set of heuristics to optimise the preallocation size
> based on the current allocation
> patterns within the file and the access patterns to the file.
> Specifying a fixed allocsize value turns off the dynamic behavior.
>
>
> Cheers
> William
>
>
> -----Ursprüngliche Nachricht-----
> Von: Laurenz Albe <[email protected]
> <mailto:[email protected]>>
> Gesendet: Mittwoch, 17. Juni 2026 13:34
> An: Raj <[email protected]
> <mailto:[email protected]>>
> Cc: Pgsql-admin <[email protected] <mailto:pgsql-
> [email protected]>>
> Betreff: Re: Size of /pgdata
>
>
>
> ACHTUNG: Diese Nachricht kommt von extern. Seien Sie kritisch
> beim Öffnen von Links und Anhängen.
>
>
>
>
> On Wed, 2026-06-17 at 11:47 +0530, Raj wrote:
> > The file count is same between them..but the file size varies.
> >
> > For example, assume node1 has issue and node2 is clean.
> >
> > 1. On both nodes file count is 76
> > 2. File size varies for some of them
> >
> > Each files are either 1GB or 2GB and some or them are 1.2gb or
> 1.5gb or between 1-2gb ....by looking at files..
> > In the problematic node, some of the 1gb files in node2 shows
> 2GB in node1...and also file size differences. making the difference.
>
> That is highly suspicious.
>
> What do you get if you run
>
> SHOW segment_size;
>
> Unless you built PostgreSQL yourself after changing the segment
> size, it should be 1GB. But then it would be impossible for some
> segment to be bigger than 1GB.
>
> Are you using some weird file system that reports file sizes wrongly?
>
> What does "ls -l" on one of these big files show?
>
> Yours,
> Laurenz Albe
>
>
>
> ________________________________
>
> Disclaimer:
>
> Diese Nachricht und ihr eventuell angehängte Dateien sind nur für
> den Adressaten bestimmt. Sie kann vertrauliche oder gesetzlich
> geschützte Daten oder Informationen beinhalten. Falls Sie diese
> Nachricht irrtümlich erreicht hat, bitten wir Sie höflich, diese
> unter Ausschluss jeglicher Reproduktion zu löschen und die
> absendende Person zu benachrichtigen. Danke für Ihre Hilfe.
>
> This message and any attached files are for the sole use of the
> recipient named above. It may contain confidential or legally
> protected data or information. If you have received this message in
> error, please delete it without making any copies whatsoever and
> notify the sender. Thank you for your assistance.
>
R U joking? Preallocation is around kilobytes, some time megabytes.
view thread (20+ messages)
reply
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: [email protected]
Cc: [email protected], [email protected], [email protected], [email protected], [email protected]
Subject: Re: AW: Size of /pgdata
In-Reply-To: <[email protected]>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox