On Fri, Dec 29, 2023 at 4:47 AM Adrian Klaver <adrian.klaver@aklaver.com> wrote:
>
> On 12/28/23 12:41, Sean Flaherty wrote:
> > I'm rather new to the mailing list, are there any additional steps I
> > should take (i.e. posting to pgsql-hackers, etc.)?
> >
>
> For what purpose? You are seeing differences in compression strategies
> between lz4 and pglz. The 'fix' would be to go back to pglz.
Agreed, lz4 is known for its high compression speed, but lower
compression ratio, this is the trade off one should bear in mind.
>
> >
> > On Thu, Dec 28, 2023 at 11:23 AM Adrian Klaver
> > <adrian.klaver@aklaver.com <mailto:adrian.klaver@aklaver.com>> wrote:
> >
> > On 12/28/23 09:13, Sean Flaherty wrote:
> > > Follow-up:
> > > Working with AWS, we found that starting in RDS Postgres 15, the
> > > default_toast_compression parameter is set to use lz4 compression
> > > instead of pglz. This resulted in the increased json storage
> > size we
> > > were seeing.
> > >
> > > I have been able to reproduce the increased storage size on RDS
> > Postgres
> > > and using my local docker instance of postgres 15.5 by changing the
> > > local default_toast_compression value in postgresql.conf.
> > >
> > > I have attached the test script we use to create a table, insert
> > some
> > > test records and a query to test the JSON data size on disk.
> >
> > I can confirm I see the same results using Postgres 16 installed from
> > the PGDG repo on Ubuntu 22.04. That the lz4 data size is greater then
> > the pglz data size.
> >
> > >
> > >
> > > Kind regards,
> > > Sean
> > >
> >
> >
> > --
> > Adrian Klaver
> > adrian.klaver@aklaver.com <mailto:adrian.klaver@aklaver.com>
> >
>
> --
> Adrian Klaver
> adrian.klaver@aklaver.com
>
>
>
--
Regards
Junwang Zhao
