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

On Tue, Apr 14, 2020 at 3:41 PM Dilip Kumar <dilipbalaut(at)gmail(dot)com> wrote:
>
>
> >
> > > @@ -281,6 +281,24 @@ DecodeXactOp(LogicalDecodingContext *ctx,
> > > XLogRecordBuffer *buf)
> > > }
> > > case XLOG_XACT_ASSIGNMENT:
> > > break;
> > > + case XLOG_XACT_INVALIDATIONS:
> > > + {
> > > + TransactionId xid;
> > > + xl_xact_invalidations *invals;
> > > +
> > > + xid = XLogRecGetXid(r);
> > > + invals = (xl_xact_invalidations *) XLogRecGetData(r);
> > > +
> > > + if (!TransactionIdIsValid(xid))
> > > + break;
> > > +
> > > + ReorderBufferAddInvalidation(reorder, xid, buf->origptr,
> > > + invals->nmsgs, invals->msgs);
> > >
> > > Why should we insert an WAL record for such cases?
> > >
> >
> > Right, if there is any such case, we should avoid it.
>
> I think we don't have any such case because we are logging at the
> command end. So I have created an assert instead of the check.
>

Have you tried to ensure this in some way? One idea could be to add
an Assert (to check if transaction id is assigned) in the new code
where you are writing WAL for this action and then run make
check-world and or make installcheck-world.

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