Intel 10Gb nic (ix driver) and bizarre netstat output

[Alexander Sack]

M

On Mon, Jun 28, 2010 at 12:37 PM, Mike Carlson <carlson39 at llnl.gov> wrote:
> I've got a 10Gb intel nic on a FreeBSD 8.0-p3/AMD64 system, using the ix
> driver:
>
> ix0: <Intel(R) PRO/10GbE PCI-Express Network Driver, Version - 1.8.9> port
> 0xdce0-0xdcff mem 0xdf3a0000-0xdf3bffff,0xdf3c0000-
> 0xdf3fffff,0xdf39c000-0xdf39ffff irq 35 at device 0.0 on pci5
> ix0: Using MSIX interrupts with 17 vectors
> ix0: [ITHREAD]
> ...
> ix0: Ethernet address: 00:1b:21:3f:b5:fc
>
> ix0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 9194
>
>  options=5bb<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,VLAN_HWCSUM,TSO4,LRO>
>    ether 00:1b:21:3f:b5:fc
>    inet 192.168.6.56 netmask 0xffffff00 broadcast 192.168.6.255
>    media: Ethernet autoselect (10Gbase-LR <full-duplex>)
>    status: active
>
> What seems a bit odd is even when they system is 'idle', netstat reports a
> burst of outgoing data at an unpredictable interval:
> # netstat -I ix0 -w 1
>            input          (ix0)           output
>   packets  errs      bytes    packets  errs      bytes colls
>         1     0        496          1     0 18446744073709551436     0
>         1     0        252          1     0          0     0
>        28     0      19768         42     0          0     0
>         2     0        316          1     0          0     0
>         1     0        252          1     0          0     0
>
> That large "burst" becomes more frequent if there is a reasonable network
> load on the server (like a NFS or Samba file transfer):
> # netstat -I ix0 -w 1
>            input          (ix0)           output
>   packets  errs      bytes    packets  errs      bytes colls
>         1     0        496          1     0 18446744073709551436     0
>      1652     0    1601979       2310     0     113716     0
>       337     0     521230        397     0 18446744073709437900     0
>      7562     0     276543      14580     0   21130649     0
>     33784     0   90243561      65289     0    6398014     0
>     34929     0  101041195      67431     0 18446744073708070746     0
>     36180     0  102019403      70112     0     932461     0
>     36337     0  104575965      70340     0 18446744073708694467     0
>     35933     0  104627291      69241     0 18446744073707935998     0
>     36498     0  104697232      70544     0 18446744073709094889     0
>     36580     0  106044621      70737     0 18446744073708270130     0
>     22934     0   80783509      44337     0 18446744073694340268     0
>     11469     0   34850586      22131     0 18446744073707453470     0
>     15661     0   40976798      30191     0    3816924     0
>     14885     0   40763491      28572     0    1794493     0
>     14554     0   47191162      28140     0 18446744073703905316     0
>     16620     0   43843276      32154     0    3277056     0
>     11800     0   38234856      22832     0 18446744073704976762     0
>     14640     0   37771926      28340     0    3720383     0
>     12230     0   37078829      23631     0 18446744073707401669     0
>
> Is this normal? Or, is there something strange about my network environment,
> and if so, are there any suggestions to help me narrow down the issue?

Have you tried running tcpdump -i ix0 to see what the 42 packets were
when its "idle?"

-aps