把網卡中斷綁定到CPU,最大化網卡的吞吐量(轉)

先來看一下問題, 我們通過 ifconfig 查看接口的名稱 為 p15p1, 一般機器為 eth0

再通過命令

➜  ~ cat /proc/interrupts | head -n 1 && cat /proc/interrupts |grep p15p1 CPU0 CPU1 CPU2 CPU3 CPU4 CPU5 CPU6 CPU7  130: 6235174 0 0 0 0 0 0 0 PCI-MSI 3145728-edge p15p1

我們看到全部的網卡中斷都集中到了 CPU0, 因此在有巨大網絡流量的時候CPU0可能過載, 讓網卡的數據包收發出現延遲. 為了分散CPU處理網卡中斷, 我們需要設把網卡綁定到特定的CPU核心. (我的服務器因為只有一個網卡, 一般2U的機架服務器都有4個網卡, 綁定到不同的CPU核心.)

中斷親和性設置是通過掩碼來標記的, 假設我們有一個8核心的CPU, 那么掩碼為 11111111, 從左到右分別為 CPU0..CPU7, 位置關系如下:

1 1 1 1 1 1 1 1 - - - - - - - - 7 6 5 4 3 2 1 0

設置中斷親和性的SHELL腳本

#!/bin/bash # # Copyright (c) 2014, Intel Corporation # # Redistribution and use in source and binary forms, with or without # modification, are permitted provided that the following conditions are met: # # * Redistributions of source code must retain the above copyright notice, # this list of conditions and the following disclaimer. # * Redistributions in binary form must reproduce the above copyright # notice, this list of conditions and the following disclaimer in the # documentation and/or other materials provided with the distribution. # * Neither the name of Intel Corporation nor the names of its contributors # may be used to endorse or promote products derived from this software # without specific prior written permission. # # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" # AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE # IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE # DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE # FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL # DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR # SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER # CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, # OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. # # Affinitize interrupts to cores # # typical usage is (as root): # set_irq_affinity -x local eth1 <eth2> <eth3> # # to get help: # set_irq_affinity usage() { echo echo "Usage: $0 [-x|-X] {all|local|remote|one|custom} [ethX] <[ethY]>" echo " options: -x Configure XPS as well as smp_affinity" echo " options: -X Disable XPS but set smp_affinity" echo " options: {remote|one} can be followed by a specific node number" echo " Ex: $0 local eth0" echo " Ex: $0 remote 1 eth0" echo " Ex: $0 custom eth0 eth1" echo " Ex: $0 0-7,16-23 eth0" echo exit 1 } usageX() { echo "options -x and -X cannot both be specified, pick one" exit 1 } if [ "$1" == "-x" ]; then XPS_ENA=1 shift fi if [ "$1" == "-X" ]; then if [ -n "$XPS_ENA" ]; then usageX fi XPS_DIS=2 shift fi if [ "$1" == -x ]; then usageX fi if [ -n "$XPS_ENA" ] && [ -n "$XPS_DIS" ]; then usageX fi if [ -z "$XPS_ENA" ]; then XPS_ENA=$XPS_DIS fi num='^[0-9]+$' # Vars AFF=$1 shift case "$AFF" in remote) [[ $1 =~ $num ]] && rnode=$1 && shift ;; one) [[ $1 =~ $num ]] && cnt=$1 && shift ;; all) ;; local) ;; custom) ;; [0-9]*) ;; -h|--help) usage ;; "") usage ;; *) IFACES=$AFF && AFF=all ;; # Backwards compat mode esac # append the interfaces listed to the string with spaces while [ "$#" -ne "0" ] ; do IFACES+=" $1" shift done # for now the user must specify interfaces if [ -z "$IFACES" ]; then usage exit 1 fi # support functions set_affinity() { VEC=$core if [ $VEC -ge 32 ] then MASK_FILL="" MASK_ZERO="00000000" let "IDX = $VEC / 32" for ((i=1; i<=$IDX;i++)) do MASK_FILL="${MASK_FILL},${MASK_ZERO}" done let "VEC -= 32 * $IDX" MASK_TMP=$((1<<$VEC)) MASK=$(printf "%X%s" $MASK_TMP $MASK_FILL) else MASK_TMP=$((1<<$VEC)) MASK=$(printf "%X" $MASK_TMP) fi printf "%s" $MASK > /proc/irq/$IRQ/smp_affinity printf "%s %d %s -> /proc/irq/$IRQ/smp_affinity\n" $IFACE $core $MASK case "$XPS_ENA" in 1) printf "%s %d %s -> /sys/class/net/%s/queues/tx-%d/xps_cpus\n" $IFACE $core $MASK $IFACE $((n-1)) printf "%s" $MASK > /sys/class/net/$IFACE/queues/tx-$((n-1))/xps_cpus ;; 2) MASK=0 printf "%s %d %s -> /sys/class/net/%s/queues/tx-%d/xps_cpus\n" $IFACE $core $MASK $IFACE $((n-1)) printf "%s" $MASK > /sys/class/net/$IFACE/queues/tx-$((n-1))/xps_cpus ;; *) esac } # Allow usage of , or - # parse_range () { RANGE=${@//,/ } RANGE=${RANGE//-/..} LIST="" for r in $RANGE; do # eval lets us use vars in {#..#} range [[ $r =~ '..' ]] && r="$(eval echo {$r})" LIST+=" $r" done echo $LIST } # Affinitize interrupts # setaff() { CORES=$(parse_range $CORES) ncores=$(echo $CORES | wc -w) n=1 # this script only supports interrupt vectors in pairs, # modification would be required to support a single Tx or Rx queue # per interrupt vector queues="${IFACE}-.*TxRx" irqs=$(grep "$queues" /proc/interrupts | cut -f1 -d:) [ -z "$irqs" ] && irqs=$(grep $IFACE /proc/interrupts | cut -f1 -d:) [ -z "$irqs" ] && irqs=$(for i in `ls -Ux /sys/class/net/$IFACE/device/msi_irqs` ;\ do grep "$i:.*TxRx" /proc/interrupts | grep -v fdir | cut -f 1 -d : ;\ done) [ -z "$irqs" ] && echo "Error: Could not find interrupts for $IFACE" echo "IFACE CORE MASK -> FILE" echo "=======================" for IRQ in $irqs; do [ "$n" -gt "$ncores" ] && n=1 j=1 # much faster than calling cut for each for i in $CORES; do [ $((j++)) -ge $n ] && break done core=$i set_affinity ((n++)) done } # now the actual useful bits of code # these next 2 lines would allow script to auto-determine interfaces #[ -z "$IFACES" ] && IFACES=$(ls /sys/class/net) #[ -z "$IFACES" ] && echo "Error: No interfaces up" && exit 1 # echo IFACES is $IFACES CORES=$(</sys/devices/system/cpu/online) [ "$CORES" ] || CORES=$(grep ^proc /proc/cpuinfo | cut -f2 -d:) # Core list for each node from sysfs node_dir=/sys/devices/system/node for i in $(ls -d $node_dir/node*); do i=${i/*node/} corelist[$i]=$(<$node_dir/node${i}/cpulist) done for IFACE in $IFACES; do # echo $IFACE being modified dev_dir=/sys/class/net/$IFACE/device [ -e $dev_dir/numa_node ] && node=$(<$dev_dir/numa_node) [ "$node" ] && [ "$node" -gt 0 ] || node=0 case "$AFF" in local) CORES=${corelist[$node]} ;; remote) [ "$rnode" ] || { [ $node -eq 0 ] && rnode=1 || rnode=0; } CORES=${corelist[$rnode]} ;; one) [ -n "$cnt" ] || cnt=0 CORES=$cnt ;; all) CORES=$CORES ;; custom) echo -n "Input cores for $IFACE (ex. 0-7,15-23): " read CORES ;; [0-9]*) CORES=$AFF ;; *) usage exit 1 ;; esac # call the worker function setaff done # check for irqbalance running IRQBALANCE_ON=`ps ax | grep -v grep | grep -q irqbalance; echo $?` if [ "$IRQBALANCE_ON" == "0" ] ; then echo " WARNING: irqbalance is running and will" echo " likely override this script's affinitization." echo " Please stop the irqbalance service and/or execute" echo " 'killall irqbalance'" fi

原地址在: https://gist.github.com/SaveT...點擊預覽

把一個網卡綁定到特定的CPU核心.

root@ubuntu:~# ./set_irq_affinity.sh one 6 p15p1 IFACE CORE MASK -> FILE ======================= p15p1 6 40 -> /proc/irq/130/smp_affinity WARNING: irqbalance is running and will likely override this script's affinitization. Please stop the irqbalance service and/or execute 'killall irqbalance'

注意輸出

IFACE CORE MASK -> FILE ======================= p15p1 6 40 -> /proc/irq/130/smp_affinity

系統默認有中斷負載均衡的進程, 如果是手工綁定, 需要停止它

killall irqbalance

查看結果

➜ ~ cat /proc/interrupts | head -n 1 && cat /proc/interrupts |grep p15p1

在一個終端中監控, 在另一個終端中設置

# 監控 watch -n 1 cat /proc/interrupts | head -n 1 && cat /proc/interrupts |grep p15p1 # 設置 ./set_irq_affinity.sh one 1 p15p1 ./set_irq_affinity.sh one 2 p15p1 ./set_irq_affinity.sh one 3 p15p1

可以看到中斷數的增長,從CPU1到CPU2,再到CPU3(注意CPU編號從0開始的, 8核的CPU編號是0-7)