(Sorry if this comes twice; I tried to send it before remembering I wasn't
subscribed to tech-net.)


At 6:14 PM -0700 9/17/99, Bill Studenmund wrote (replying to me):

>> My NetBSD box is a Motorola StarMax 5000/250 with a pair of NIC cards in
>> it, both using the Digital chip. de0 is connected to a DSL bridge. de1 is
>> connected to a hub, and the system routes for the rest of my network.
>>
>> Most of the time, it works great.  But if I'm putting a constant load on
>> the network (not a heavy load, just a constant one), I get periodic
>> "pauses" -- every 15-60 minutes, I'll suddenly get a 15-30 second period
>> during which I can't ping the router at my ISP.  Then it returns to normal.
>> I've tried swapping de0 out for another card, with no effect.  The ISP
>> can't find any problems, nor can the DSL line provider.  The only clue I
>> have is this error in my messages log, which appears several times whenever
>> the pause occurs:
>>
>> Sep 15 19:42:14 ghek /netbsd: arplookup: unable to enter address for
>> 204.85.32.5 (host is not on local network)
>>
>> (204.85.32.5 is not a local machine; it's the host I'm communicating with
>> when I generate the constant traffic.  Okay, I'm playing Warbirds.  I admit
>> it.  But the problem's definitely not at their end; I can't ping any
>> addresses at all beyond de0's IP.)
>>
>> Does this suggest anything to anyone?  Any ideas what the problem might be?
>
>Bring this up on tech-net. Something's messing up with arp. You might also
>try ping -n whent hings are bad as ping will normally trigger a name
>lookup.

So, I'm bringing it up on tech-net.  I've done a lot more troubleshooting
since.

First, I know the problem's not at the ISP's end.  When I hook my desktop
mac up to the DSL line directly, there are no problems.  Also, although
I've duplicated the problem by setting up a Centris 650 with netbsd/mac68k
to do the routing and getting the same results as I do with the StarMax, I
can use the same Centris 650 running Mac OS with a program called
IPNetRouter to do the routing, and the problem does not occur.

Oddly enough, the problem was also present with the StarMax running Yellow
Dog Linux.

I've done a tcpdump on both interfaces just today.  The problem seems to
occur here:

11:27:51.829725 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
207.126.1
01.100 tell 207.207.64.141
11:27:51.829729 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
207.126.1
01.100 tell 207.207.64.141
11:27:51.829734 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
207.126.1
01.100 tell 207.207.64.141
11:27:51.829738 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
207.126.1
01.100 tell 207.207.64.141
11:27:51.829743 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
207.126.1
01.100 tell 207.207.64.141
11:27:51.829747 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
204.85.32
.13 tell 207.207.64.141
11:27:51.829752 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
204.85.33
.5 tell 207.207.64.141
11:27:51.829756 0:0:94:82:9d:74 ff:ff:ff:ff:ff:ff 0806 42: arp who-has
207.207.6
4.142 tell 207.207.64.141

It corresponds with this in the messages log:

Oct  3 11:27:13 ghek /netbsd: arplookup: unable to enter address for
207.126.101
.100 (host is not on local network)
Oct  3 11:27:21 ghek last message repeated 4 times
Oct  3 11:27:22 ghek /netbsd: arplookup: unable to enter address for
204.85.32.1
3 (host is not on local network)
Oct  3 11:27:23 ghek /netbsd: arplookup: unable to enter address for
204.85.33.5
 (host is not on local network)

Not sure why the timestamp doesn't match.

Any ideas?

Thanks,

-Bill


--
 William R. Dickson -- Consuming the Earth's limited resources since 1968
 wrd@awenet.com   < >   http://www.awenet.com/~wrd/   < >   ICQ: 14117523

     Death: You're an interesting young man. We'll meet again.
     Boris: Don't bother.
     Death: It's no bother.
                                       - Woody Allen,
                                         Love and Death