arp problem Solaris 10

george

unread,

Jan 20, 2009, 8:27:49 AM1/20/09

to

A bit of a riduculous problem . There are 2 servers on the same subnet. If
I run a ping -s <target> from one of the servers ( as a network test) I see
that every 40 minutes exactly there is a hang of about 18 seconds. During
this period the mac address of the <targer> disappers from the arp cache as
seen by "arp -a" . A snoop of the network shows the client sending an arp
request for the <target> server which does not seem to be answered .

Eddie

unread,

Jan 20, 2009, 9:49:15 AM1/20/09

to

What type of hardware are you on, what type of network card is
installed, more info would be useful for other to help. Is this a
T2000 by chance?

george

unread,

Jan 20, 2009, 10:08:43 AM1/20/09

to

# prtdiag -v | more
System Configuration: Sun Microsystems sun4u Sun Fire V245
System clock frequency: 188 MHZ
Memory size: 8GB

# ndd /dev/ip ip_ire_arp_interval ### Looks like 20 minutes
1200000

I wrote a script to run arp -a and grep the targets ip address to detect
when the entry disappears from arp cache and it works out at exactly 40
minutes, almost to the second : ie
no arp entry at Tuesday, January 20, 2009 2:14:22 PM GMT
no arp entry at Tuesday, January 20, 2009 2:54:21 PM GMT

Here is the prtconf -v output for the network cards :
pci 188 MB pci14e4,1668 (network)
okay /pci@1e,600000/pci@0/pci@9/pci@0/network@4

pci 188 MB pci14e4,1668 (network)
okay /pci@1e,600000/pci@0/pci@9/pci@0/network

pci 188 MB pci14e4,1668 (network)
okay /pci@1e,600000/pci@0/pci@a/pci@0/network

George

"Eddie" <edval...@gmail.com> wrote in message
news:2a8cd35a-3eec-4c21...@a26g2000prf.googlegroups.com...

george

unread,

Jan 20, 2009, 10:52:44 AM1/20/09

to

I did another snoop. I can see that the target server ,instead of replying
with a the mac address of the clilent issuing the arp request it is itself
sending an arp request to the client :

em01 -> (broadcast) ARP C Who is 10.200.054.156, f245h.ziggy.com ?
em01 -> (broadcast) ARP C Who is 10.200.054.156, f245h.ziggy.com ?
f245h.ziggy.com -> (broadcast) ARP C Who is 10.200.054.152, em01 ?
em01 -> f245h.ziggy.com ARP R 10.200.054.152, em01 is
0:14:4f:a3:aa:3c

George

"george" <geo...@twig.tk> wrote in message
news:AKadncGUWb2ddOjU...@posted.plusnet...

Colin B.

unread,

Jan 20, 2009, 12:29:21 PM1/20/09

to

Check Sunsolve for arp patches. I don't have the bugID handy, but this
sounds VERY similar to something we ran in to a year ago with Solaris 9
and 10 both.

If I come across the patch/bugID, I'll post it.

Colin

Martha Starkey

unread,

Jan 20, 2009, 1:25:49 PM1/20/09

to

Could this be the infamous "broadcom arp" issue first reported in Dec 07
after S10 Update 4 was introduced? You want to rule this out and the
question is "are there Dell teamed mode NICS on that subnet?"

References:

http://blogs.sun.com/swas/entry/solaris_10_8_07_broadcom

http://blogs.sun.com/swas/entry/update_to_the_broadcom_pc

There's a Sun Spectrum doc on this as well if you have access: Problem
Resolution # 236964

george

unread,

Jan 22, 2009, 10:07:32 AM1/22/09

to

The solution to this problem of a server not responding to arp requests was
remarkably simple.

The ifconfig command has an arp option
ipconfig -arp , then ifconfig arp.
a console has to be used because "ipconfig -arp" may flush the arp table
leaving the server uncontactable until "ifconfig arp" is run

George