> On 19/01/2023 01:23 CET Ian Lawrence Barwick <barwick@gmail.com> wrote:
>
> 2023年1月19日(木) 8:50 Erik Wienhold <ewie@ewie.name>:
> >
> > > On 19/01/2023 00:09 CET Hilbert, Karin <ioh1@psu.edu> wrote:
> > >
> > > I manage some PostgreSQL clusters on Linux. We have a Primary & two Standby
> > > servers & for Production, there is also a DR server. We use repmgr for our HA
> > > solution & the Standbys are cloned from the Primary using the repmgr standby
> > > clone command.
> > >
> > > My manager asked for a report of all the user databases & their sizes for each
> > > server in the cluster. I used the psql "\l+" command & then extracted the
> > > database name & the size from the output. I expected the databases to be the
> > > same size on the Standbys as on the Primary, but I found that some of the
> > > databases were smaller on the Standby servers than on the Primary.
> > >
> > > For example, the output on the Primary for one of the user databases showed
> > > as: 8997 kB, but on the Standbys, it was 8849 kB.
> >
> > The standbys could be missing some indexes because schema changes are not
> > replicated and must be applied manually.
>
> This is incorrect; with streaming replication all changes applied on the primary
> are applied on the standby.
Thanks. I was thinking about logical replication.
> (...)
> > Could also be different page sizes. But that's a compilation option. What does
> > SHOW block_size say on those systems?
>
> It is impossible to start a standby using binaries built with a
> different block size to
> the primary.
Makes sense for streaming replication.
--
Erik