On Mon, Aug 10, 2020 at 10:24 AM Masahiko Sawada
<masahiko.sawada@2ndquadrant.com> wrote:
>
> It's true that heap_page_is_all_visible() is called from only
> lazy_vacuum_page() but I'm concerned it would lead misleading since
> it's not actually removing tuples but just checking after vacuum. I
> guess that the errcontext should show what the process is actually
> doing and therefore help the investigation, so I thought VACUUM_HEAP
> might not be appropriate for this case. But I see also your point.
> Other vacuum error context phases match with vacuum progress
> information phrases. So in heap_page_is_all_visible (), I agree it's
> better to update the offset number and keep the phase VACUUM_HEAP
> rather than do nothing.
>
Okay, I have changed accordingly and this means that the offset will
be displayed for the vacuum phase as well. Apart from this, I have
fixed all the comments raised by me in the attached patch. One thing
we need to think is do we want to set offset during heap_page_prune
when called from lazy_scan_heap? I think the code path for
heap_prune_chain is quite deep, so an error can occur in that path.
What do you think?
--
With Regards,
Amit Kapila.