Situation: Two Red Hat 6.1/6.2 boxes, connected via Cat5 to a hub. One
box (felix) has a modem connected to it, and dials out to the internet.
The other box (calvin) wishes to surf via felix. I did some preliminary
tests....
1. I can ping from felix (directly connected to the internet).
2. I can ping felix from calvin, and vice versa.
3. At first, I couldn't even ping the internet from calvin. Or so it
seemed.
4. Checking tcpdump on felix's ppp0 interface, I could see that the
pings from calvin to the internet were going out and being answered.
However, they weren't showing up on calvin.
5. Forwarding maybe? So I set up an ipchains rule, to wit:
ipchains -a forward -s 192.168.0.0/24 -d 0/0 -j MASQ
This is the _only_ firewall rule at this point, and the default policy
for all the chains is ACCEPT. (Don't whine, I'll fix that after I get
this problem solved.)
6. So now I get ping answers on calvin, but the waits are bizarre. First
ping is sub-second. Second is a minute, third is about 90 seconds,
fourth is about 2-1/2 minutes. I can tell this because ping -v shows the
sequence numbers of the pings.
7. Now, oddly enough, monitoring that same sequence of pings on the ppp0
interface of felix with tcpdump, the pings come back rapid-fire, one per
second. So they're going out and coming in properly from felix, but
calvin isn't getting the responses in the same time frame.
8. So maybe LAN collisions? Nope. Checked ifconfig, and there were no
collisions on the ppp0 interface, and only two on the eth0 interface.
9. FWIW, /etc/nsswitch.conf shows "hosts: files dns". /etc/resolv.conf
shows two nameservers, in the standard "nameserver xxx.xxx.xxx.xxx"
configuration. /etc/hosts shows all the hosts on the network properly.
So I'm baffled. Why would pings travel via the gateway (felix) at speed,
return at speed, but take minutes to go back to the originating machine
from the gateway?
Paul
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 14:15:59 EDT