perfSONAR User Q&A and Other Discussion

["Lang, Michael" <>]

Thanks Eli.

 

I tried updating firmware but it looks like the only firmware available is boot ROMs.  Updating that didn’t help.

 

Next I tried updating the driver to the latest (4.1.2) and that has fixed the problem.  Something must be wrong with the way ixgbe that ships in the kernel with my system.

 

Can anyone tell me whether running a stock intel ixgbe driver rather than the web100 is going to cause any issues with perfSONAR functionality?

 

-          Mike

 

From: Eli Dart [mailto:]
Sent: Wednesday, August 12, 2015 6:23 PM
To: Lang, Michael <>
Cc:
Subject: Re: [perfsonar-user] Issues with Intel 10G NIC, X520

 

I haven't seen this specifically, but this sort of issue looks like it might be a driver or firmware bug.

 

Are you able to change the version of the ixgbe driver you are using, or run different firmware on the card?  (I forget if those cards have separate firmware).

 

Eli

 

 

On Wed, Aug 12, 2015 at 10:01 AM, Lang, Michael <> wrote:

Hello,

 

I’ve been having trouble with a system running perfSONAR, specifically with the 10G NIC.  The system is a Dell PowerEdge R610 with an Intel X520 NIC installed into it.  The problem that I’m having is that with newer kernels there has been a consistent load of .5 on the system.  After troubleshooting this I found the load is likely from the kernel thread ‘events’ getting stuck in state ‘D’ (Uninterruptible sleep) often.  Looks like this from top:

 

   21 root      20   0     0    0    0 D  0.0  0.0   6:42.19 events/2

 

Another symptom (the one I first noticed) is that keystrokes are choppy through SSH – especially if events/0 is the one going to ‘D’.  If I pull the ixgbe driver from the system these problems go away. 

 

The last kernel that I had working that didn’t have this problem was:

 

2.6.32-504.16.2.el6.web100

 

The current kernel installed is:

 

2.6.32-573.1.1.el6.web100.x86_64

 

I’ve tried changing some available ixgbe parameters such as interrupts and flow director but it hasn’t helped.  The only parameter I’m currently running is to use unsupported transceivers, but I don’t believe that should cause this issue.

 

Has anyone else seen this?  Anyone know of a fix or anything I could try to help resolve it?

 

Thanks,

 

-          Mike

 

--

Eli Dart, Network Engineer                          NOC: (510) 486-7600

ESnet Office of the CTO (AS293)                          (800) 333-7638

Lawrence Berkeley National Laboratory 

PGP Key fingerprint = C970 F8D3 CFDD 8FFF 5486 343A 2D31 4478 5F82 B2B3