https://bugs.openldap.org/show_bug.cgi?id=8174

--- Comment #5 from Howard Chu hyc@openldap.org --- (In reply to Vladimir Cunat from comment #4)

FYI, blocking the transaction after mdb_drop isn't great. Some projects got caught by that: postfix, knot-dns, knot-resolver (I didn't search specifically, so maybe more).

Personally I find it useful to clear the whole DB and fill new contents, all within a single RW transaction (so that noone sees the WIP state during that rewrite). But I guess everyone will have to try to find some workarounds now...

This particular behavior probably should be reverted, will look at it for v1.0.1.

Affected applications should stick to LMDB v0.9 in the meantime.