This is a known issue.
It stems from systemd not being able to cope properly with legacy init.d boot scripts (*) and it starts the nfs server before it starts rpc and therefore nfs server doesn’t work.
You can either add a delay (sleep 20) to the init.d script for nfs-kernel or you can add a sleep/restart nfs statement to /etc/rc.local/ by adding the following lines just above exit(0)
sleep 20
systemctl restart nfs-common
(*) Many would argue that this is the other way round, but who’s fault it is is rather a moot point: nfs-server is fundamentally broken in debian whoeve’s fault it is.