Hello all FW-friends,
First of all thanks for everybody who sent me
experiences about this.
I read some material that confirms the TCP timeout
problem but the strange thing in my case is that is occurs in UDP
traffic...
I like the SP3 idea, and I'll check wheter it is
installed or not and I'll keep in touch,
Once again, thanks very much for all those who sent
me suggestions!
BTW, here is the summary of
suggestions:
>Aylton, > >this is a known
problem with TCP sessions that have >timed out. > >SP3 fixes this problem
>I
had a problem with this. After installing an Alteon Web >Switch.
We had to increase the time-out that it held a >connection open before
closing it (in the Alteon). I was >seeing lots and lots of packets
leave through the firewall >and not get natted. These were all the tail
ends of closing >TCP sessions. Once we made the change in the Alteon
>switch, the number of these incidents per day dropped >down to below 50
per day. From up in the many hundreds >per day.
>I have this
problem too. I think that in a service pack of >FW 4.1 >version it's
supposed to be fixed..Since I upgraded to >4.1 sp3 it >happens less
times, but still it's present.
>Enrypted packets (FWZ, SKIP, IKE) cannot have NAT
>applied, as their TCP/IP headers are encrypted, however >IPSEC only
encrypts the data portion, leaving the headers >free for NAT
manipulation.
>Also bear in mind that NAT won't work with protocols that
>use embedded IP addresses (RPC, Oracle etc), unless a >Proxy has been
written for them
Daniel
>A
little more detail on this:
>
>You CAN use NAT on encrypted packets using FWZ,
>SKIP, and IKE with AH only (no ESP; see below) since >the packets are not
encapsulated - the original header will >be translated. You can also
use NAT on IKE packets >using ESP for encapsulation, but the NAT will apply
prior >to the encryption/encapsulation. This allows you to do
>things like create NAT rules to pass traffic over a VPN to >sites with
the same addressing on both ends.
>Not sure if this answers Aylton's original
question, but >hopefully worth at least
$0.02.
|