Re: PATCH: logical_work_mem and logical streaming of large in-progress transactions

On Wed, Apr 8, 2020 at 6:29 AM Tomas Vondra
<tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
>
> On Tue, Apr 07, 2020 at 12:17:44PM +0530, Amit Kapila wrote:
> >On Mon, Mar 30, 2020 at 8:58 PM Tomas Vondra
> ><tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
> >
> >I think having something like we discussed or what you have in the
> >patch won't be sufficient to clean the KnownAssignedXid array. The
> >point is that we won't write a WAL for xid-subxid association for
> >unlogged relations in the "Immediately WAL-log assignments" patch,
> >however, the KnownAssignedXid would have both kinds of Xids as we
> >autofill it with gaps (see RecordKnownAssignedTransactionIds). I
> >think if my understanding is correct to make it work we might need
> >major surgery in the code or have to maintain KnownAssignedXid array
> >differently.
>
> Hmm, that's a good point. If I understand correctly, the issue is
> that if we create new subxact, write something into an unlogged table,
> and then create new subxact, the XID of the first subxact will be "known
> assigned" but we won't know it's a subxact or to which parent xact it
> belongs (because there will be no WAL records that could encode it).
>

Yeah, there could be multiple such missing subxacts.

> I wonder if there's a simple solution (e.g. when creating the second
> subxact we might notice the xid-subxid assignment was not logged, and
> write some "dummy" WAL record).
>

That WAL record can have multiple xids.

> But I admit it seems a bit ugly.
>

Yeah, I guess it could be tricky as well because while assembling some
WAL record, we need to generate an additional dummy record or might
need to add additional information to the current record being formed.
I think the handling of such WAL records during hot-standby and in
logical decoding could vary. During logical decoding, currently, we
don't form an association for subtransactions if it doesn't have any
changes (see ReorderBufferCommitChild) and now with this new type of
record, I think we need to ensure that we don't form such association.

I think after quite some changes, tweaks and a lot of testing, we
might be able to remove XLOG_XACT_ASSIGNMENT but I am not sure if it
is worth doing along with this patch. I think it would have been good
to do this if we are adding any visible overhead with this patch and
or it is easy to do that. However, none of that seems to be true, so
it might be better to write good comments in the code indicating what
all we need to do to remove XLOG_XACT_ASSIGNMENT so that if we feel it
is important to do in future we can do so. I am not against spending
effort on this but I don't see the urgency of doing it along with this
patch.

--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com