Re: Is Recovery actually paused?

On Sun, Jan 24, 2021 at 12:16 PM Bharath Rupireddy
<bharath(dot)rupireddyforpostgres(at)gmail(dot)com> wrote:
>
> On Sun, Jan 24, 2021 at 7:17 AM Dilip Kumar <dilipbalaut(at)gmail(dot)com> wrote:
> > On Sat, 23 Jan 2021 at 4:40 PM, Bharath Rupireddy <bharath(dot)rupireddyforpostgres(at)gmail(dot)com> wrote:
> >>
> >> On Sat, Jan 23, 2021 at 1:36 PM Dilip Kumar <dilipbalaut(at)gmail(dot)com> wrote:
> >> > Please find the patch for the same. I haven't added a test case for
> >> > this yet. I mean we can write a test case to pause the recovery and
> >> > get the status. But I am not sure that we can really write a reliable
> >> > test case for 'pause requested' and 'paused'.
> >>
> >> +1 to just show the recovery pause state in the output of
> >> pg_is_wal_replay_paused. But, should the function name
> >> "pg_is_wal_replay_paused" be something like
> >> "pg_get_wal_replay_pause_state" or some other? To me, when "is" exists
> >> in a function, I expect a boolean output. Others may have better
> >> thoughts.
> >>
> >> IIUC the above change, ensuring the recovery is paused after it's
> >> requested lies with the user. IMHO, the main problem we are trying to
> >> solve is not met
> >
> >
> > Basically earlier their was no way for the user yo know whether the recovery is actually paused or not because it was always returning true after pause requested. Now, we will return whether pause requested or actually paused. So > for tool designer who want to wait for recovery to get paused can have a loop and wait until the recovery state reaches to paused. That will give a better control.
>
> I get it and I agree to have that change. My point was whether we can
> have a new function pg_wal_replay_pause_and_wait that waits until
> recovery is actually paused ((along with pg_is_wal_replay_paused
> returning the actual state than a true/false) so that tool developers
> don't need to have the waiting code outside, if at all they care about
> it? Others may have better thoughts than me.

I think the previous patch was based on that idea where we thought
that we can pass an argument to pg_is_wal_replay_paused which can
decide whether to wait or return without the wait. I think this
version looks better to me where we give the status instead of
waiting. I am not sure whether we want another version of
pg_wal_replay_pause which will wait for actually it to get paused. I
mean there is always a scope to include the functionality in the
database which can be achieved by the tool, but this patch was trying
to solve the problem that there was no way to know the status so I
think returning the correct status should be the scope of this.

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