Our V880 locked up for no apparent reason a couple of days ago, where 
locked up means it would not respond on the network or the serial 
console and had to be reset.  There was nothing in the system logs to 
indicate a problem, nor were there any power events logged on the UPS 
(or by other machines in the same room), prtdiag showed the hardware it 
tested to be ok, and SMART analysis of the disks showed no issues.

So today the vts diagnostics are being run.  Everything passed except 
nettest, which said this for eri0:

05/29/08 10:27:36 gec SunVTS5.1ps13: VTSID 6002 nettest.
   ERROR eri0(/pci@9,700000/network@1,1): "No ICMP echo reply from 
131.215.XX.YY."
   Probable_Cause(s):
     (1)bad network board
     (2)system load too heavy
     (3)no cable connection
     (4)target machine too busy
   Recommended_Actions:
     (1)replace network board
     (2)reduce system load or increase timeout time
     (3)check cable connection
     (4)reduce target machine load

We can rule out causes 2,3,4 .  If this is a bad network board, it seems 
to only want to fail for nettest. Despite what nettest says, the network 
seems to be working just fine on the V880 (for instance, I ran vts 
remotely over that very network).  Nettest chose XX.YY to be one less 
than the address of the V880, and that happened to be a printer.  That 
printer pings just fine not only from other machines, but from the V880.

% ping -I 1 131.215.XX.YY

Does not drop a single packet in 100 tries.  Thinking that nettest might 
send the packets faster than 1 per second, I tried it from a linux host 
using

ping -i .01 131.215.XX.YY

and the printer was able to return pings that fast.  (Ping on solaris 
did not allow this test.)  netstat -i shows a couple of errors:

Name  Mtu  Net/Dest      Address        Ipkts  Ierrs Opkts  Oerrs Collis 
Queue
lo0   8232 loopback      localhost      14562131 0     14562131 0     0 
      0
eri0  1500 gec           gec            785446 0     723980 3     0 
  0

There were also a couple of funky messages in /var/adm/messages after 
the reboot yesterday that I don't recall seeing before:

May 28 11:27:31 gec inetd[187]: [ID 965992 daemon.error] 100229/1-2/tcp: 
unknown service
May 28 11:27:31 gec inetd[187]: [ID 965992 daemon.error] 100230/1/tcp: 
unknown service
May 28 11:28:33 gec inetd[187]: [ID 965992 daemon.error] 100229/1-2/tcp: 
unknown service
May 28 11:28:33 gec inetd[187]: [ID 965992 daemon.error] 100230/1/tcp: 
unknown service

Is nettest right and the network board is going, or is nettest itself 
messed up?  I don't have a network loopback connector, so cannot run 
netlbtest.

Opinions?

Thanks,

David Mathog

David Mathog wrote:
> Is nettest right and the network board is going, or is nettest itself 
> messed up?  I don't have a network loopback connector, so cannot run 
> netlbtest.

Made a network loopback connector (two actually, since I had to cut up a 
patch cable to do it) and then ran netlbtest for a long time, during 
which it logged no errors.  So back to the regular nettest and 
configured it to target the local network switch, instead of the printer 
it picked by itself.  Ran nettest 25 times, no errors.  My best guess is 
that even though the printer can normally handle a high rate of pings, 
perhaps when the errors occurred it was accepting a print job, or 
performing some other task, and so could not keep up with nettest.

Also, for future reference, in order to get the serial console to work
properly with "sunvts -t" through TeraTerm (or probably any other VT100 
emulator) it was necessary to first do:

export TERM=vt100

otherwise TERM defaulted to "sun" and "sunvts -t" resulted in an 
unusable mess of an interface.

Regards,

David Mathog

"David Mathog" <mathog@caltech.edu> wrote in message
news:g1mqp5$1ee$1@naig.caltech.edu...
> Our V880 locked up for no apparent reason a couple of days ago, where
> locked up means it would not respond on the network or the serial
> console and had to be reset.  There was nothing in the system logs to
> indicate a problem, nor were there any power events logged on the UPS
> (or by other machines in the same room), prtdiag showed the hardware it
> tested to be ok, and SMART analysis of the disks showed no issues.

Probably not what you want to hear, but it could be the motherboard is bad.
Especially if when it locks up there is no way to send a break and collect a
core file.

All the CPU/Memory boards plug into this and the ethernet is onboard.

Depending on the P/N or rev level of your motherboard there are some known
issues with the onboard ethernet.
The info is here:

http://www.sunsolve.sun.com/search/document.do?assetkey=I1195

although access requires a SunSolve account.

Hopefully you have a Sun service contract.
I have no idea what such a part is worth, but even on eBay people are asking
a lot.

Trinean

Trinean wrote:
> "David Mathog" <mathog@caltech.edu> wrote in message
> news:g1mqp5$1ee$1@naig.caltech.edu...
>> Our V880 locked up for no apparent reason a couple of days ago, where
>> locked up means it would not respond on the network or the serial
>> console and had to be reset.  There was nothing in the system logs to
>> indicate a problem, nor were there any power events logged on the UPS
>> (or by other machines in the same room), prtdiag showed the hardware it
>> tested to be ok, and SMART analysis of the disks showed no issues.
> 
> Probably not what you want to hear, but it could be the motherboard is bad.
> Especially if when it locks up there is no way to send a break and collect a
> core file.

Yeah, I wasn't happy that it wouldn't respond to a break.  If the system 
does it again fairly soon I'll have to put in a service call.  In the 
meantime the ethernet  (re)tested clean.

> Depending on the P/N or rev level of your motherboard there are some known
> issues with the onboard ethernet.
> The info is here:
> 
> http://www.sunsolve.sun.com/search/document.do?assetkey=I1195

Using an account linked to a service contract I still couldn't follow 
that link.  Is the link current?

Thanks,

David Mathog