Dear all,
I'm new to this list, if this isn't the right place to post this issue,
please let me know.
I am testing two Intel 82598EB 10-Gigabit AT network card mounted on two
Dell R710 server with Red Hat Enterprise Linux 5.4 installed. The two
NICs are directly connected each other (no switch in between). I am
experiencing a problem on the receive side of the connection: I'm sending
an UDP stream after having disabled flow control on each card, and about
the 60% of the transmitted packets are dropped with a large number of
rx_missed_errors. This is what ifconfig and ethtool -S show:
[r...@grids2]# ethtool -S eth0
NIC statistics:
rx_packets: 1761794
tx_packets: 32
rx_bytes: 14748667944
tx_bytes: 5746
lsc_int: 2
tx_busy: 0
non_eop_descs: 6510828
rx_errors: 0
tx_errors: 0
rx_dropped: 0
tx_dropped: 0
multicast: 2
broadcast: 2
rx_no_buffer_count: 0
collisions: 0
rx_over_errors: 0
rx_crc_errors: 0
rx_frame_errors: 0
rx_fifo_errors: 0
rx_missed_errors: 1148382
tx_aborted_errors: 0
tx_carrier_errors: 0
tx_fifo_errors: 0
tx_heartbeat_errors: 0
tx_timeout_count: 0
tx_restart_queue: 0
rx_long_length_errors: 0
rx_short_length_errors: 0
tx_tcp4_seg_ctxt: 0
tx_tcp6_seg_ctxt: 0
tx_flow_control_xon: 0
rx_flow_control_xon: 0
tx_flow_control_xoff: 0
rx_flow_control_xoff: 0
rx_csum_offload_good: 28
rx_csum_offload_errors: 0
tx_csum_offload_ctxt: 11
low_latency_interrupt: 0
alloc_rx_page_failed: 0
alloc_rx_buff_failed: 0
lro_aggregated: 8
lro_flushed: 8
lro_recycled: 0
rx_no_dma_resources: 0
hw_rsc_count: 0
rx_flm: 0
fdir_match: 0
fdir_miss: 0
tx_queue_0_packets: 32
tx_queue_0_bytes: 5746
rx_queue_0_packets: 2
rx_queue_0_bytes: 120
rx_queue_1_packets: 2
rx_queue_1_bytes: 120
rx_queue_2_packets: 1761762
rx_queue_2_bytes: 14748664800
rx_queue_3_packets: 0
rx_queue_3_bytes: 0
rx_queue_4_packets: 0
rx_queue_4_bytes: 0
rx_queue_5_packets: 14
rx_queue_5_bytes: 1452
rx_queue_6_packets: 7
rx_queue_6_bytes: 726
rx_queue_7_packets: 7
rx_queue_7_bytes: 726
[r...@grids2]# ifconfig eth0
eth0 Link encap:Ethernet HWaddr 00:1B:21:4B:C8:BF
inet addr:10.100.100.2 Bcast:10.100.100.255 Mask:255.255.255.0
inet6 addr: fe80::21b:21ff:fe4b:c8bf/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:9000 Metric:1
RX packets:1761795 errors:0 dropped:1148382 overruns:0 frame:0
TX packets:39 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:14748668004 (13.7 GiB) TX bytes:9332 (9.1 KiB)
Each interface has the MTU set to 9000, the driver is the one provided
by Dell support (ixgbe 2.0.44.14-NAPI) with Rx multiqueue enabled and
InterruptThrottleRate set to 8000 (but I tried 16000 and 32000 too).
Also I tried to set FdirPballoc parameter to 0, 1 and 2 with no success.
Single CPU load is no more than 10% with 8000 interrupt/s:
[r...@grids2 ~]# mpstat -P ALL 2
05:16:43 PM CPU %user %nice %sys %iowait %irq %soft %steal
%idle intr/s
05:16:45 PM all 0.00 0.00 0.06 0.00 0.06 1.19 0.00
98.69 9018.00
05:16:45 PM 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.00 1002.50
05:16:45 PM 1 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.00 1.50
05:16:45 PM 2 0.00 0.00 0.50 0.00 0.50 9.45 0.00
89.55 8007.00
05:16:45 PM 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.00 1.00
05:16:45 PM 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.00 1.00
05:16:45 PM 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.00 3.00
05:16:45 PM 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.00 1.00
05:16:45 PM 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.00 1.00
I tried to configure smp affinity to assign the IRQ of each
rx-queue to a different CPU, with no effect on the amount of dropped
packets. I also verified that if I connect the NIC on a 10GE switch and
send more than one stream from different servers, the card uses
different rx-queues but the problem still exists.
I modified the /etc/sysctl.conf file parameters adding:
net.core.wmem_max = 16777216
net.core.wmem_default = 8388608
net.core.rmem_max = 16777216
net.core.rmem_default = 8388608
net.ipv4.tcp_rmem = 4096 262144 16777216
net.ipv4.tcp_wmem = 4096 262144 16777216
net.ipv4.tcp_mem = 4096 8388608 16777216
net.core.optmem_max = 524288
net.core.netdev_max_backlog = 200000
Some other details:
- CPU: dual quad core Intel L5520 @ 2.27GHz (Hyper-Threading disabled)
- RAM: 24 GB (12 X 2GB)
- Slot PCI: PCIe gen.2 x8
- Kernel linux: 2.6.18-128.1.1.el5 and 2.6.18-164.2.1.el5
Any suggestions?
Thank you very much in advance.
Mirko Corosu
--
--------------------------------------------------------------------------
Mirko Corosu
Network and system administrator
Computing Center
Istituto Nazionale Fisica Nucleare
Via Dodecaneso 33
16146 Genova, Italy
http://www.ge.infn.it
--------------------------------------------------------------------------
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit
http://communities.intel.com/community/wired
