On Tue, Jul 12, 2022 at 5:45 AM Andres Freund <andres@anarazel.de> wrote:
> Hm - given that we've observed ftruncate failing with strace, and that
> stracing to find problems isn't insane, shouldn't we retry the ftruncate too?
> It's kind of obsoleted by your next patch, but not really, because it's not
> unconceivable that other OSs behave similarly? And IIUC you're planning to not
> backpatch 0002?
Yeah. Done, and pushed. 0002 not back-patched.
> > + pgstat_report_wait_start(WAIT_EVENT_DSM_FILL_ZERO_WRITE);
>
> (not new in this patch, just moved around) - FILL_ZERO_WRITE is imo an odd
> description of what's happening... In my understanding this isn't doing any
> writing, just allocating. But whatever...
True. Fixed in master.