Re: Parallel scan with SubTransGetTopmostTransaction assert coredump

On Tue, Aug 3, 2021 at 11:41 PM Greg Nancarrow <gregn4422(at)gmail(dot)com> wrote:
> I've tried to follow your description and have attached a patch to
> hopefully match it, but it doesn't pass "make check-world".
> Perhaps I messed something up (apologies if so), or additional changes
> are needed to match what you had in mind or correct additional issues
> you didn't foresee?

This is the sort of thing I was thinking about but I don't understand
why it doesn't fix the reported problem. Apparently I haven't
correctly understood what the issue is.

> >I don't know why Greg's patch is changing anything related to the
> >active snapshot (as opposed to the transaction snapshot). Maybe
> >there's a reason why we need that change, but I don't know what it is.
>
> I don't think my v2/v5 patch is changing anything related to the
> active snapshot (is it?).
> It's restoring the serialized active snapshot, installing it as the
> transaction snapshot and setting it as the active snapshot.

Why do you think it's right to install the serialized *active*
snapshot as the *transaction* snapshot? I've been operating on the
presumption that we wanted the worker to install the leader's
transaction snapshot as its transaction snapshot as its transaction
snapshot and the leader's active snapshot as its active snapshot,
because in my mind the active snapshot and the transaction snapshot
are two separate things. However, there's clearly something wrong with
that line of reasoning, because apparently your proposal fixes the
problem and mine doesn't.

--
Robert Haas
EDB: http://www.enterprisedb.com