Re: Error handling in after-startup shmem requests

Hi,

On Sat, 29 Aug 2026 at 02:35, Andres Freund <andres(at)anarazel(dot)de> wrote:

> Hi,
>
> On 2026-08-28 17:58:55 +0300, Heikki Linnakangas wrote:
> > On 26/08/2026 15:44, Ayush Tiwari wrote:
> > > I took a quick look and triggered the tests parallely.
> > >
> > > Patch looks good to me.
> >
> > Committed, thanks!
>
> This test reliably fails for me locally. I think it may be due to my system
> having huge pages configured and the default of the huge_pages GUC being
> try. When using huge pages, we round up the allocation, which increases the
> reserve. Which then triggers the test to fail to fail, as the reserve is
> bigger than 128kB.
>

Thanks for the report.

The attached patch disables huge pages immediately before the allocation-
failure phase. This keeps the test focused on recovery from an allocation
failure rather than on the platform's shared memory segment rounding.

Regards,
Ayush