Previous 1 2 Next 18 Replies Latest reply: Feb 20, 2015 11:10 AM by ntDog
colbruA Level 1 Level 1

Hi

 

I've just updated 3 MacPro's 3.1 and 4 MacPro's 1.1 from 10.5.8 to SL 10.6.8.

 

All the Macs have been runing fine until SL was installed.

Now all couple of hours the Macs are bringing the Ethernet Link down and up again. (not togeter it's random)

 

Here from the Kernel.log:

Nov 16 04:56:41 su09-23 kernel[0]: Ethernet [Intel8254X]: Link down on en0

Nov 16 04:56:44 SU09-23 kernel[0]: Ethernet [Intel8254x]: Link up on en0, 1-Gigabit, Full-duplex, No flow-control, Debug [792d,ac08,0de1,0e00,c1e1,7800]

 

It cannot be the switch! First it used to work for years with this untouched switch and cabling.

Second even if I leave out the switch and connect a MacPro directly to a server it still happens.

 

My guess is that the Drivers for 10.6.8 are buggy. Becaue the HW is fine. It can't be that all the HW is now all of the sudden broken.

And btw I have just bought brand new MacPros 5.1 and they do not drop the link. And they are running 10.6.8 too.

But they of course use another network chip and driver.

 

So anyone can confirm this or even has a solution for this mess?

 

 

Thanks for your contibution.

 

Bruno


Mac OS X (10.6.8)
  • Linc Davis Level 10 Level 10
    expertise.applications
    Applications

    I see those messages sometimes, and my Ethernet is working fine. Is yours?

  • colbruA Level 1 Level 1

    No it's not working fine. I have iSCSI mounted Voumes and they lose the connection. The iSCSI Server also reports that the Link went down on the client.

  • Linc Davis Level 10 Level 10
    expertise.applications
    Applications

    Any packet losses when you ping the initiators?

  • colbruA Level 1 Level 1

    No not at all. the communication is fast and stable. It's just the 1.1 and 3.1 Macs that lose the link randomly.

    I also have one 4.1 and that one has no problem too as the 5.1 MacPro's

  • Linc Davis Level 10 Level 10
    expertise.applications
    Applications

    The log messages are normal. I get the same messages. The problem lies in the iSCSI system. I suggest you take it up with the vendor.

  • colbruA Level 1 Level 1

    I don't agree. Why would the newer MacPros then have no problem it the issue lies within the iSCSI System?

  • Linc Davis Level 10 Level 10
    expertise.applications
    Applications

    I don't know. If anyone knows,the SAN vendor would. All I can tell you is that the log messages are not a manifestation of a problem. Maybe if you identify the product, someone else who uses it would be able to tell you something more helpful.

  • colbruA Level 1 Level 1

    Ok I've found a workaround. I've bought PCIe Network cards with a different cipset (realtek). Installed the Cards installed the driver. And I can now capture video without interruption. The link is not going down!

    So there IS a problem in the Network chip or driver of the 1.1 and 3.1 MacPros.

  • simonjoliver Level 1 Level 1

    I have a MacPro 3,1 running 10.5.8 Server (Leopard, Not SL) - but a similar sounding problem. Periodically my logs show link being dropped for a few seconds and re-established.

     

    Also, I see about 3%-5% packet loss pinging the machine - and not only at times when the log entries show loss of link. Sometimes pings go unanswered for a second - more often 3-6 seconds; occasionally as much as 25 seconds. This is happening seemingly randomly, every minute or so. I've tried different cables, different switch ports, and both built-in Ethernet ports. Nothing makes a difference.

     

    I'm considering trying a PCIe NIC instead - but was wondering if anyone had seen this problem, and then had it go away with a more recent system upgrade. I'm wondering if taking the pain of installing Lion and rebuilding things is going to fix anything...

  • Grant Bennet-Alder Level 9 Level 9

    en0, 1-Gigabit, Full-duplex, No flow-control,

    Also consider enabling flow control.

  • colbruA Level 1 Level 1

    I can confirm that I had the very same problem as simonjoliver. OS upgrades or clean installs did not help.

    What helped was buying and installing a PCIe NIC.

  • Grant Bennet-Alder Level 9 Level 9

    Also, I see about 3%-5% packet loss pinging the machine

    That could be an indication that the physical port is going bad. There should be no measurable losses on a cable under 100 meters length.

  • The hatter Level 9 Level 9

    What about the cable? Guess worth swapping that out to see if a new cable makes any difference, right?

  • simonjoliver Level 1 Level 1

    I agree that it there should be zero packet loss.... all my other hosts have zero, including hosts on the same switch. In my case I tried different cables, different switch ports, both built-in NICs and manually configuring the Ethernet settings with and without flow control. The behaviour doesn't change.

     

    The machine is 3 years old now. So yes, it could be a hardware failure somewhere high enough up the chain to affect both interfaces; but as originally noted by colbruA it seems suspiciously like a low-level driver thing.

     

    In general the machine continues working fine - the network drops are generally brief enough to recover from - so if it was a desktop, you might not notice any problems. However, because I'm using it to serve an Ajax app that gets some brief bursts of intense connection activity, its causing noticeable problems.

     

    I have a PCIe NIC on order; I am hoping that will behave better.

Previous 1 2 Next