Howard Chu @ Symas389DS has switched their default backend from #BerkeleyDB to #LMDB. https://www.port389.org/docs/389ds/FAQ/Berkeley-DB-deprecation.html
The difference in config complexity is quite dramatic.
This bit of monitoring info is puzzling. Since LMDB read txns are lockfree/waitfree, there shouldn't be any waits. This waitingROtxn monitoring value implies they're using an additional lock system on top of LMDB, which guarantees they will have poor performance, particularly on multicore servers.
That project still seems pretty pointless, even after leveraging all of OpenLDAP's ultra-efficient libraries.
