Hi, all:
I am having an interesting problem I have been unable to solve following
my
standard methods. Perhaps you can help me out.
I have a Mandrake 6.1 machine with a 2.2.14 kernel that I configured in
March to run telnet and ftp
services. After months of trouble free service it is now only
intermittently
accepting connections. This machine
worked fine up until late last week. It is not behind a firewall.
I have been telnetting in from both an NT box and a Linux box behind a
firewall. Again, this
has worked fine for about three months.
Now, when I try to log in using this same telnet and firewall
configuration, my telnet clients inform me that the connection failed.
I cannot ftp either.
Now, here is where it gets bizarre -
When I telnet into my mail server (a separate machine, also running
Mandrake 6.1, on the same subnet as the machine that doesn't work)
from behind the firewall, I can log in just
fine -- now, when I telnet over to my failing box, I can log in without
trouble!
Perhaps it's a DNS issue, huh? Then explain this:
Once I have done this, it becomes possible to log into the failing box
using the NT client behind a firewall.
So I can't log in until I log in from the neighbouring Mandrake box --
once I've done this, I *can* log in -- and suddenly ftp works again,
too.
Now, if I log out from the neighbouring Mandrake machine so that only
the
NT machine is logged in, I can keep working for a short time, then *paf*
-
it disconnects me. Once again it is not possible to log in again.
Now, there may have been some idle time that caused this, but the point
is
that it won't let me log in after booting me off -- unless I go through
the rigamarole of logging in over my Mandrake box first.
I thought it might be tcpd, but the logs show no evidence of a connect
attempt, even to inetd. It looks like inetd is on crack or something.
Only
successful logins show up in any of the logs.
Then I thought it might be the firewall:
I cannot connect from a Linux box behind the firewall - but I also know
of
another Windows user who was having the same problems, and that person
was
not behind a firewall.
The hostname on the firewall router is correct -- and it was working up
until last week (there haven't been any modifications to it).
I had one more theory that I tested - I thought it might be apmd, so I
killed it. Things worked fine for a few hours, and then the problem was
back.
The following things *have* changed in the last two weeks:
The firewall router was rebooted (that doesn't explain why people not
behind a firewall have the same problem)
There was a power failure in the room housing the two Mandrake boxes
(they
are not UPS'ed, but I haven't found any evidence of damage yet - they
rebooted on their own without trouble, and they are surge protected)
Since encountering the problem I have updated my netkit-base to
netkit-base-0.16-1mdk, with no luck. I used tcpchk to make sure my tcp
configuration wasn't on the blink.
This problem is very peculiar indeed and I would love some help with it
=)
Thanks,
Stephen Bosch
