lmdb maturity

Hi folks,

Would anyone want to comment on the maturity of lmdb?

I've been mostly swayed by the compelling technical arguments for it for most purposes of an instance of OpenLDAP, by whitepapers and benchmarks. I've yet to read many real complaints/known major issues against it.

But still today, the manual pages for hdb and lmdb state hdb remains the preferred normal use database backend. There's also the future file [in]compatibility bit in the slapd-mdb man page.

Any thoughts on when lmdb may be declared "good enough for production loads" and/or the default recommended backend?

Thanks!