On Tue, 24 Mar 2020 at 13:53, Amit Kapila <amit.kapila16@gmail.com> wrote:
>
> On Tue, Mar 24, 2020 at 9:46 AM Justin Pryzby <pryzby@telsasoft.com> wrote:
> >
> > On Mon, Mar 23, 2020 at 02:25:14PM +0530, Amit Kapila wrote:
> > > > Yea, and it would be misleading if we reported "while scanning block..of
> > > > relation" if we actually failed while writing its FSM.
> > > >
> > > > My previous patches did this:
> > > >
> > > > + case VACUUM_ERRCB_PHASE_VACUUM_FSM:
> > > > + errcontext("while vacuuming free space map of relation \"%s.%s\"",
> > > > + cbarg->relnamespace, cbarg->relname);
> > > > + break;
> > > >
> > >
> > > In what kind of errors will this help?
> >
> > If there's an I/O error on an _fsm file, for one.
> >
>
> If there is a read or write failure, then we give error like below
> which already has required information.
> ereport(ERROR,
> (errcode_for_file_access(),
> errmsg("could not read block %u in file \"%s\": %m",
> blocknum, FilePathName(v->mdfd_vfd))));
Yeah, you're right. We, however, cannot see that the error happened
while recording freespace or while vacuuming freespace map but perhaps
we can see the situation by seeing the error message in most cases. So
I'm okay with the current set of phases.
I'll also test the current version of patch today.
Regards,
--
Masahiko Sawada http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services