On Wed, May 25, 2022 at 3:55 AM Matthias van de Meent < boekewurm+postg...@gmail.com> wrote:
> On Wed, 25 May 2022 at 10:15, Zhihong Yu <z...@yugabyte.com> wrote: > > > > Hi, > > Patch v3 follows advice from Matthias (status field has been dropped). > > Could you argue why you think that this should be added to the > pg_stat_progress_copy view? Again, the progress reporting subsystem is > built to "report the progress of certain commands during command > execution". Why do you think we need to go further than that and allow > some commands to retain their report even after they've finished > executing? > > Of note: The contents of >st_progress_param are only defined and > guaranteed to be consistent when the reporting command is running. > Even if no other progress-reporting command is running other commands > or processes in that backend may call functions that update the fields > with somewhat arbitrary values when no progress-reporting command is > actively running, thus corrupting the information for the progress > reporting view. > > Could you please provide some insights on why you think that we should > change the progress reporting guts to accomodate something that it was > not built for? > > > Kind regards, > > Matthias van de Meent > Hi, Matthias: When I first followed the procedure in https://paquier.xyz/postgresql-2/postgres-14-monitoring-copy/ , I didn't see the output from the view. This was because the example used 10 rows where the COPY command finishes quickly. I had to increase the row count in order to see output from the system view. With my patch, the user would be able to see the result of COPY command even if the duration for command execution is very short. I made a slight change in patch v4. With patch v3, we would see the following: relid | command | status_yb | type | bytes_processed | bytes_total | tuples_processed | tuples_excluded -------+-----------+-----------+------+-----------------+-------------+------------------+----------------- - | COPY FROM | PASS | PIPE | 6 | 0 | 1 | 0 It would be desirable to see the relation for the COPY command. With the updated patch, I think the interference from other commands in progress reporting has been prevented (see logic inside pg_stat_get_progress_info). Cheers
v4-copy-progress-status.patch
Description: Binary data
