On Tue, Mar 03, 2020 at 10:05:42PM +0900, Masahiko Sawada wrote:
> I was concerned about fsm vacuum; vacuum error context might show heap
> scan while actually doing fsm vacuum. But perhaps we can update
> callback args for that. That would be helpful for user to distinguish
> that the problem seems to be either in heap vacuum or in fsm vacuum.
On Tue, Mar 03, 2020 at 04:49:00PM -0300, Alvaro Herrera wrote:
> Your use of the progress-report enum now has two warts -- the "-1"
> value, and this one,
>
> > +#define PROGRESS_VACUUM_PHASE_VACUUM_FSM 7 /* For error reporting only */
>
> I'd rather you define a new enum, in lazyvacuum.c.
On Mon, Mar 16, 2020 at 11:44:25AM +0530, Amit Kapila wrote:
> > On Wed, Mar 04, 2020 at 04:21:06PM +0900, Masahiko Sawada wrote:
> > > Thank you for updating the patch. But we have two more places where we
> > > do fsm vacuum.
> >
> > Oops, thanks.
...
> it is not clear whether it is a good idea to keep a phase like
> VACUUM_ERRCB_PHASE_VACUUM_FSM as it has added additional updates in
> multiple places in the code.
I think you're suggesting to rip out VACUUM_ERRCB_PHASE_VACUUM_FSM, and allow
reporting any errors there with an error context like "while scanning heap".
An alternative in the three places using VACUUM_ERRCB_PHASE_VACUUM_FSM is to
set:
|phase = VACUUM_ERRCB_PHASE_UNKNOWN;
to avoid reporting any error context until another phase is set.
--
Justin