Another weird systemd problem (I guess) : stunnel

[Tom Horsley]

I run stunnel as a service using an old sysv style
service init.d script I cobbled up a while back. I
checked to make sure it dynamically creates the
/var/run/stunnel directory it needs since /var/run
is now tempfs, and it already had code to do that.

The service appears to start, and I can even see
stunnel processes running, but if I try to actually
use the stunnel connection, it doesn't work, for
example:

postfix/smtpd[2568]: lost connection after CONNECT from localhost[127.0.0.1]

That is postfix trying to use stunnel to talk to
a relayhost to send mail right after a boot.

Weirdly, if I do "service stunnel restart", I can
then use stunnel with no problems.

I finally put this line in rc.local, and stunnel
seems to be working now when I boot:

/bin/bash -c 'sleep 5 ; service stunnel restart' > /dev/null 2>&1 < /dev/null &

I can't think what might not be working at the time
the stunnel service originally starts, since it is
very late in boot.log, but maybe it has the same
problem folks have been reporting with sendmail and
NFS mounts?