According to Tom Keffer there are two strategies.
I can not use the init.d script "weewx-multi" since I am using fedora which does not have "/lib/init/vars.sh" or "/lib/lsb/init-functions" since it uses systemd (and systemctl).
What I did is I just modified the standard "/etc/init.d/weewx" script to "/etc/init.d/weewx_atlas"
I just changed "NAME=weewx_atlas" for the first instance.
Then changed "NAME=weewx_wmod" for the second instance and
"NAME=weewx_merge" for the third instance of the weewx demon.
Then I need three files /etc/weewx/weewx_atlas.conf and weewx_wmod.conf and weewx_merge.conf.
The database file names and HTML_ROOTs have to be unique in each.
Other than that the main difference is the weewx_merge.conf has under [DATA_BINDINGS] two sections [atlas_binding] with "database=archive_sqlite_atlas" and [wmod_binding] with "database=archive_sqlite_wmod". And under [Databases] two sections [archive_squlite_wmod] with "database_name = weewx_wmod.sdb" and [archive_sqlite_atlas] with "database_name = weewx_atlas.sdb".
Everything else is same but ...
[[SeasonsReport]]
[[[ImageGenerator]]]
[[[[day_images]]]]
[[[[[daybarometer]]]]]
[[[[[[wmodBarometer]]]]]]
data_type = barometer
data_binding = wmod_binding
label = WM barometer
[[[[[[atlasBarometer]]]]]]
data_type = barometer
data_binding = atlas_binding
# label = AT barometer
label = AT barom
and so on for all sensors
So the question is how to prevent pipeline from attempting to do archive operations and change the databases? And what driver to use? I do not really need a driver.