CentOS - Jan 2005 - crossposting: OpenVPN on CentOS(RHES clone) and Fedora Core3..some strange things

Apologies for the cross-posting to CentOS list (already mailed to OpenVPNlist)..

am using OpenVPN [ http://openvpn.sf.net ] on CentOS 3.4 and having
some issues..
it may possibly be an OpenVPN issue, but thought if anyone on this
list is using or had any
similar issues or have any thoughts..

TIA

---------- Forwarded message ----------
Date: Wed, 12 Jan 2005 11:26:22 -0700
Subject: resend with updates: OpenVPN on CentOS(RHES clone) and Fedora
Core3..some strange things
To: Openvpn-users at lists.sourceforge.net


below is my original posting with the issue I have been having..
couple of updates (alas, I still have the problem/issue)..please
help..anyone..anyone..thanx

1. rebuilt the CentOS box/applied latest patches..(clone of RHES 3.0)

2. snagged 2.0 rc7 openvpn and compiled (this time pointing to already
centos installed openssl RPMs/files)
http://sourceforge.net/mailarchive/forum.php?thread_id=6328352&forum_id=8453
this time, I even generated the server certs on CentOS itself..

3. same issue, OpenVPN doesn't respond aftrer in production for a
day..I don't see any log activity
that stands out..note, I am using TCP instead of UDP..

4. only reboot is fixing this issue..

5. anyone successfully using OpenVPN on RedhatEnterprise 3.x or any
clones of it with out any
issue?
TIA

====my original posting to the list====
Greetings folks,

env:
1. servers on OVPN 2.0 rc6 on FC3 and CentOS 3.3(clone of RHES 3.0 with updt3),
with all latest updates..Clients are all XP SP2. using TUN on servers/clients..
2. using in routed config, custom certs all generated on FC3 and
copied/replicated on to centos box.
3.  use diff  ip/netwok for "server" (in server config file) directive
for obvious reasons..both OVPN servers are serving the same
corp/office network..
4. using TCP instead of UDP.
5. same setof "iptables" firewall rulesets..

here is the strange thing..

OVPN on FC3 seems to work with out any issues, but the one on CentOS
box does not respond,
after in production for a while..I have to reboot for it to start
responding..my verbose logs don't show
anything..it's like, there is not even a connection being
established..I have ruled out network issues etc..
external IP's on both hosts are valid and accessable, while OpenVPN on
443 port gets hung or doesn't respond on CentOS host...(i run it on
TCP 443)...

am getting a little tired of rebooting centos box. i guess I could
cron the reboot, but I would like to
find the root cause..anyone seen this issue?
TIA
============================

On Wed, 12 Jan 2005, j p wrote:
> Apologies for the cross-posting to CentOS list (already 
> mailed to OpenVPNlist)..
> 1. rebuilt the CentOS box/applied latest patches..(clone of RHES 3.0)
Why rebuild? -- Also, there is not likely to be anything about 
Fedora Core3 addressed here
> 2. snagged 2.0 rc7 openvpn and compiled (this time pointing to already
Why the latest rather than a prior stable if you are having 
problems?  Basic debugging practice is to alter one variable 
at a time from the last known working configuration.
> 3. same issue, OpenVPN doesn't respond aftrer in production for a
> day..I don't see any log activity
> that stands out..note, I am using TCP instead of UDP..
Why?  UDP is more tested
> 4. only reboot is fixing this issue..
What does high logging under openvpn, or a tcpdump trace 
listing for traffic only on your communication port show?
> 5. anyone successfully using OpenVPN on RedhatEnterprise 3.x or any
> clones of it with out any
I use OpenVPN extensively, on Centos-3 series hosts, and have 
no problems.  It may be useful to build from the files at:
     ftp://ftp.owlriver.com/pub/mirror/ORC/openvpn/
  - and -
     ftp://ftp.owlriver.com/pub/mirror/ORC/lzo/

which are known to produce a stable and functional VPN mesh

Good luck

-- Russ Herrold