Johan Ymerson wrote:
Heiko Hund <hh...@astaro.com> wrote on 2010-04-07 15:26:05:
Then I don't think we see the same problem. We have been using 2.0.9 on
~100 machines for a couple of years now without these problems. We
upgraded 3 of them to 2.1.1, and they all instantly got this problem. We
downgraded them to 2.0.9, and all problems gone. Then started switching
between 2.0.9 and 2.1_rcX to determine when this problem started, but all
2.1_rcX seems to suffer from it, but 2.0.9 does _not_! So something must
have changed between 2.0.9 and 2.1 that at least makes this a lot worse.
Strange, we tested with 2.1_rc22 and the driver from 2.0.9 and ran into the
problem within ten attempts. I can't image that the problem is related to
userspace openvpn. Do you have the problem every time you connect and do you
see the DNS servers in the output of `ipconfig /all`?
I have not tested with 2.1.1 and tap driver from 2.0.9. Vanilla 2.0.9 does not
have this problem (at least not in the thousands of connect attempts I have
made by now). Vanilla 2.1.1 (and all rc versions I have found) does have the
problem. It shows up in 1-2 attempts.
`ipconfig /all` shows all the correct values.
And, this is funny, nslookup always work.
what exactly do you mean with 'nslookup always works' ?
I bet this actually is a change in the behavior of the userspace part, not the
driver. Something userspace did after configuring the interface, but it doesn't
do anymore. Something that made Windows re-examine the DNS order.
this could be a timing issue, which could be caused by changes to the
userspace openvpn part; did you try playing with the
route-delay
tap-sleep
dhcp-renew
flags ? did that have any affect?
cheers,
JJK
