
Hi, I installed mailman3 a few years ago on lists.sandelman.ca aka lists.tcpdump.org.
My upgrade was in part a response to the HTTP based attacks that used the signup form on mailman2 to attack people via SMTP with DOS. I know the IETF mailman2 install {now gone} had some kind of delay implemented in the web form before you could submit. I didn't understand why. I installed via debian packages, and I'd really prefer to stay with that. (Not what many maintainers would say, I know. Essential that it work if mailman3 is going to take over) There were problems, probably with the hyperkitty integration with subsequent updates seemed to solve.
I installed with sqlite3, since it seemed if all work was going to go through the mailman3 daemon, that it would handle all the concurrency issues. This has not been the case. I get weekly complaints about the database being locked. Sometimes this goes way on it's own, but sometimes it does not.
A good way to tickle this is to go the web interface, go to a list with 200 held messages, and then having told it to show me 100+ messages per page, to then tell it to discard them all. So slow. Before I realized what was going on, I'd open multiple tabs to try to make this concurrent.
So I tried to switch last week from sqlite3 to postgresql. I used pgloader with a configuration file attached below. This seemed easier than a sqlite3 dump/psql, as the SQL requires some massaging that seeemed to require more brainpower than I had at the time. I then edited /etc/mailman3/mailman.cfg to suit. I had set up the postgresql database under the "list" owner, so unix-domain socket permissions, and restarted.
I then saw some errors: lists not found! I visited the web interface. NO LISTS. DAMN. I put it back. I haven't figured out what went wrong yet. I will clone the VM and try again. I will report whatever I find here.
=== load database from '/var/lib/mailman3/mailman.db' into postgresql:///mailman3
with include drop, create tables, create indexes, reset sequences
set work_mem to '16MB', maintenance_work_mem to '512 MB';
ps: I'd pay some $$ money for email support consulting.