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

On Sat, Jan 4, 2020 at 10:00 AM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>
> On Sun, Dec 29, 2019 at 1:34 PM Dilip Kumar <dilipbalaut(at)gmail(dot)com> wrote:
> > On Sat, Dec 28, 2019 at 9:33 PM Tomas Vondra
> > <tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
> > +static void
> > +set_schema_sent_in_streamed_txn(RelationSyncEntry *entry, TransactionId xid)
> > +{
> > + MemoryContext oldctx;
> > +
> > + oldctx = MemoryContextSwitchTo(CacheMemoryContext);
> > +
> > + entry->streamed_txns = lappend_int(entry->streamed_txns, xid);
> > +
> > + MemoryContextSwitchTo(oldctx);
> > +}
> > I was looking into the schema tracking solution and I have one
> > question, Shouldn't we remove the topxid from the list if the
> > (sub)transaction is aborted? because once it is aborted we need to
> > resent the schema.
> >
>
> I think you are right because, at abort, the subscriber would remove
> the changes (for a subtransaction) including the schema changes sent
> and then it won't be able to understand the subsequent changes sent by
> the publisher. Won't we need to remove xid from the list at commit
> time as well, otherwise, the list will keep on growing.
Yes, we need to remove the xid from the list at the time of commit as well.

One more
> thing, we need to search the list of all the relations in the local
> map to find xid being aborted/committed, right? If so, won't it be
> costly doing at each transaction abort/commit?
Yeah, if multiple concurrent transactions operate on the common
relations then the list can grow longer. I am not sure how many
concurrent large transactions are possible maybe it won't be huge that
searching will be very costly. Otherwise, we can maintain the sorted
array of the xids and do a binary search or we can maintain hash?

--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com