把网卡中断绑定到CPU,最大化网卡的吞吐量

摘要：先来看一下问题我们通过查看接口的名称为一般机器为再通过命令我们看到全部的网卡中断都集中到了因此在有巨大网络流量的时候可能过载让网卡的数据包收发出现延迟为了分散处理网卡中断我们需要设把网卡绑定到特定的核心我的服务器因为只有一个网卡一般的

先来看一下问题, 我们通过 ifconfig 查看接口的名称 为 p15p1, 一般机器为 eth0

再通过命令

➜  ~ 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

我们看到全部的网卡中断都集中到了 CPU0, 因此在有巨大网络流量的时候CPU0可能过载, 让网卡的数据包收发出现延迟. 为了分散CPU处理网卡中断, 我们需要设把网卡绑定到特定的CPU核心. (我的服务器因为只有一个网卡, 一般2U的机架服务器都有4个网卡, 绑定到不同的CPU核心.)

中断亲和性设置是通过掩码来标记的, 假设我们有一个8核心的CPU, 那么掩码为 11111111, 从左到右分别为 CPU0..CPU7, 位置关系如下:

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

设置中断亲和性的SHELL脚本

#!/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  
#
# 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
" $IFACE $core $MASK
    case "$XPS_ENA" in
    1)
        printf "%s %d %s -> /sys/class/net/%s/queues/tx-%d/xps_cpus
" $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
" $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=$(
原地址在: https://gist.github.com/SaveT...
把一个网卡绑定到特定的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"
注意输出
IFACE CORE MASK -> FILE
=======================
p15p1 6    40   -> /proc/irq/130/smp_affinity
系统默认有中断负载均衡的进程, 如果是手工绑定, 需要停止它
killall irqbalance
查看结果
➜  ~ cat /proc/interrupts | head -n 1 && cat /proc/interrupts |grep p15p1
在一个终端中监控, 在另一个终端中设置
# 监控
watch -n 1 cat /proc/interrupts | head -n 1 && cat /proc/interrupts |grep p15p1
# 设置
./set_irq_affinity.sh one 1 p15p1
./set_irq_affinity.sh one 2 p15p1
./set_irq_affinity.sh one 3 p15p1
可以看到中断数的增长,从CPU1到CPU2,再到CPU3(注意CPU编号从0开始的, 8核的CPU编号是0-7)