Well... truth be told this isn't a summary, but I thought some of
you people out there might find it useful. If I'm telling people
something that they already know, okay. If not, then this caused DEC
support to scratch their heads for a while to sort out so I can
imagine someone else may get burnt.
Situation:
----------
System is an AlphaStation 4/233 running Digital UNIX v4.0f (recently
upgraded from v4.0e). It became a requirement that the workstation
should connect to the Internet and the only way this could be done
was over a PPP link via an ISP. The system is connected to an
internal network and needs that network connection to be permanently
in place (the internal network is an RFC1918 network).
Problem:
--------
Getting PPP to log into the ISP with dynamic IP addressing was done
without too much difficulty. This worked with some fiddling with the
routing tables.
When we came to dial up again later in the day, the connection was
established, but DNS lookups failed, telnet, ftp, ttp, etc all were
not working. Looking at the routing tables, these appeared to be
fine, but the interface definition for ppp0 was incorrect. Looking
at the /etc/ppp/ppp-log, the IP addresses were being dynamically
assigned correctly, but no packets were being sent over the PPP link
(using pppstats). The system was even rebooted (thinking that the
routing tables had got shot). Still no good.
Analysis:
---------
What we found happening was the kernel wasn't clearing down the
interface settings on ppp0. Therefore, when the next PPP call was
being made, the IP addresses were being added to the interface as
alias addresses. The routing table information was being created
correctly in that the ppp0 route was set to the dynamic IP address,
and the default route pointed to the PPP interface. The only thing
that was wrong was the interface IP address not being cleared
between calls.
Solution:
---------
Although not a perfect solution for many, and has caused some trouble
for us, the following method works.
First stop all routing. Send a SIGKILL to gated or routed and clean
up the routing tables (route flush or manually remove entries where
appropriate). Once this is done add any required static routes into
the routing table and /etc/routes (we are lucky in that we do not
need a lot of routing information and static routes are acceptable).
What is important here is that you DO NOT put in any default route.
This has caused our routing table to grow from about five or six
entries (when running gated) to about 15 static routes -- excluding
the loopback addresses. There may be a better way to do this, but
currently this works for us (any suggestions about a better way to
setup the routing would be MORE than greatly received).
Purge down the ppp0 interface of all previous addresses:
ifconfig ppp0 delete
When we entered this command originally, we had THREE IP addresses
purged from this interface -- this was the source of the real
trouble.
Now edit the /etc/ppp/ip-down script to include the line:
ifconfig ppp0 delete
This line removes the dynamic IP address for the PPP adapter and
removes the entry from the routing tables every time you dial out.
Hope this helps.
-jon.
--
Jon Morgan <jmorgan_at_dircon.co.uk>
DEC Systems Specialist
JRI Europe Ltd
nihil illegitemi carborvndvm
[INHERIT ('SYS$LIBRARY:DISCLAIMER.PEN')]
____________________
Received on Wed May 26 1999 - 14:41:22 NZST