Just to reply for completeness.
MWM was high enough (4GB) .
Apparently we were having max_worker_processes = 18 and 16 subscriptions running, leaving no room for additional processes. What an oversight.
Once mwp was bumped (to 42 on 64 core machine), everything started working as expected.
On Sat, Apr 25, 2020 at 5:03 AM Peter Geoghegan <
pg@bowt.ie> wrote:
On Fri, Apr 24, 2020 at 7:32 AM Radoslav Nedyalkov <rnedyalkov@gmail.com> wrote:
> We 're rebuilding a big table which has set parallel_workers = 6
> system has
> max_parallel_maintenance_workers | 6 | /var/lib/pgsql/11/data/postgresql.sumup.conf
> max_parallel_workers | 16 | /var/lib/pgsql/11/data/postgresql.sumup.conf
> max_parallel_workers_per_gather | 4 | /var/lib/pgsql/11/data/postgresql.sumup.conf
>
> Also session level on index restore there is
> set max_parallel_maintenance_workers = 6;
>
> Still we get only 2 parallel processes in a free of any other load system.
> It is postgres 11.7
Try increasing maintenance_work_mem from the default of 64MB. MWM
constrains the number of parallel workers used.
--
Peter Geoghegan