You are right, the sequence number of the reply packet is 0000 instead
of 0004 (btw. it is part of the header not the data).
OK, so this is a bug in the Zyxcel Router but should have nothing to do
with the actual network problem we have. This is our main problem:
From time to time (almost always when we work on the wire i.e. plug off
a host or plug in a new one) in our Class C net it happens that some
hosts in the same subnet become unreachable. I.e. a ping or telnet or
whatever from host1 to host2 succeeds before the change but does no
longer succeed after the change. We oberved that sometimes one or more
of these actions help to settle the situation but it is not
- reset the hub
- reboot a machine (not necessarily the machine host1 or host2)
- switch off/on a CISCO 760 ISDN Router and do a reboot
A second observation we make:
The CISCO router is reachable from most hosts in the same subnet. When
we switch off the CISCO and then on again and then reboot it, the router
becomes reachable from those few hosts it was not reachable before but
is no longer reachable from the other hosts it was reachable before.
This also happens after we plug in into the same subnet a new host for
example. Here again some nondeterminism is at work: sometimes a switch
on without a reboot is sufficient, sometimes not. sometimes an
additional reset of the hub is necessary. sometimes we have to wait for
some time (10 minutes) until it becomes reachable at all. Also these two
set of hosts are not static, i.e. a host that was not able to reach the
CISCO before the event do not reach it after the event but perhaps a few
- broken network card
- bad hub (but we already changed it with another one of the same type -
both cheap ones)
- broken CISCO
Can anybody give us a hint what it actually may be or how we can test it
further? Are there Network Monitors that may give hep us trace it down?
The net contains 10BaseT (most), 100BaseT and 3 hosts with BNC linked
with 2 ATI Hubs and 1 very cheap "level one" hub.
Barry Margolin schrieb:
> >Strange things are happening in our network. The most irritating
> >observation is this:
> >host1 (Solaris 2.5.1), host2 (Solaris 2.6) and host3 (Zyxcel Router) are
> >in the same subnet:
> >on host1: 'ping host3' returns a 'no answer from host3'. But 'telnet
> >host3' returns a proper telnet prompt
> >on host2: tcpdump shows both the echo request from host1 to host3 and
> >and the echo reply from host3 to host1
> >Even on host1 itself (the host that does not succeed with the ping)
> >tcpdump shows the ping echo reply packet.
> >What could be the reason for this behaviour?
> Does the data portion of the Echo Reply match the data in the Echo Request?
> Ping puts sequence numbers in the data portion so that it can determine
> which replies go with which requests. If the router isn't copying the data
> from the request to the reply (as required by the ICMP specification) then
> this could have the results you're seeing.
> You can use the -x option to tcpdump to see a full hex dump of the packets.
> Genuity, Burlington, MA
> *** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
> Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.