On Tue, May 31, 2016 at 8:51 PM, Peter Eisentraut
<peter.eisentraut@2ndquadrant.com> wrote:
> On 5/31/16 4:04 PM, Tom Lane wrote:
>> The name should be closely related to what we use for #3. I could go for
>> max_total_parallel_workers for #2 and max_parallel_workers for #3.
>> Or maybe max_parallel_workers_total?
>
> Most cluster-wide settings like this are named max_something
> (max_connections, max_wal_senders, max_replication_slots), whereas things
> that apply on a lower level are named max_something_per_something
> (max_files_per_process, max_locks_per_transations).
>
> So let's leave max_worker_processes mostly alone and not add any _total_,
> _absolute_, _altogether_. ;-)
That's interesting, because it suggests that max_parallel_degree might
end up being called something that doesn't begin with "max". Which is
an interesting line of thought. Now, it does function as a maximum of
sorts, but that doesn't necessarily imply that it has to have max in
the name. By way of analogy, work_mem is not called max_work_mem, yet
everybody still understands that the actual memory used might be less
than the configured value.
Now, this case is a little trickier. If we called it simply
parallel_degree rather than max_parallel_degree, then it would have
the same name as the reloption. But the reloption sets an exact
value, and the GUC sets a cap, which is a significant difference.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company