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)
文章版权归作者所有,未经允许请勿转载,若此文章存在违规行为,您可以联系管理员删除。
转载请注明本文地址:https://www.ucloud.cn/yun/9416.html
摘要:网卡将数据包通过的方式写入到指定的内存地址,该地址由网卡驱动分配并初始化。调用协议栈相应的函数,将数据包交给协议栈处理。 本文将介绍在Linux系统中,数据包是如何一步一步从网卡传到进程手中的。 如果英文没有问题,强烈建议阅读后面参考里的两篇文章,里面介绍的更详细。 本文只讨论以太网的物理网卡,不涉及虚拟设备,并且以一个UDP包的接收过程作为示例. 本示例里列出的函数调用关系来自于ke...
摘要:网卡将数据包通过的方式写入到指定的内存地址,该地址由网卡驱动分配并初始化。调用协议栈相应的函数,将数据包交给协议栈处理。 本文将介绍在Linux系统中,数据包是如何一步一步从网卡传到进程手中的。 如果英文没有问题,强烈建议阅读后面参考里的两篇文章,里面介绍的更详细。 本文只讨论以太网的物理网卡,不涉及虚拟设备,并且以一个UDP包的接收过程作为示例. 本示例里列出的函数调用关系来自于ke...
摘要:继上一篇介绍了数据包的接收过程后,本文将介绍在系统中,数据包是如何一步一步从应用程序到网卡并最终发送出去的。该函数会根据路由表和目的,找到这个数据包应该从哪个设备发送出去,如果该没有绑定源,该函数还会根据路由表找到一个最合适的源给它。 继上一篇介绍了数据包的接收过程后,本文将介绍在Linux系统中,数据包是如何一步一步从应用程序到网卡并最终发送出去的。 如果英文没有问题,强烈建议阅读后...
摘要:继上一篇介绍了数据包的接收过程后,本文将介绍在系统中,数据包是如何一步一步从应用程序到网卡并最终发送出去的。该函数会根据路由表和目的,找到这个数据包应该从哪个设备发送出去,如果该没有绑定源,该函数还会根据路由表找到一个最合适的源给它。 继上一篇介绍了数据包的接收过程后,本文将介绍在Linux系统中,数据包是如何一步一步从应用程序到网卡并最终发送出去的。 如果英文没有问题,强烈建议阅读后...
摘要:无独有偶,国内云服务商进一步迭代了负载均衡产品,成功地提升了单机性能。年之后,出现了新的网络负载均衡技术。的硬件负载均衡产品又分单机和集群系列。首先,我们来考察一下负载均衡服务器变化场景。 Maglev是谷歌为自己的数据中心研发的解决方案,并于2008开始用于生产环境。在第十三届网络系统设计与实现USENIX研讨会(NSDI ‘16)上, 来自谷歌、加州大学洛杉矶分校、SpaceX公...
阅读 2408·2021-09-26 09:47
阅读 351·2019-08-30 15:55
阅读 831·2019-08-29 15:35
阅读 1081·2019-08-29 13:48
阅读 1756·2019-08-26 13:29
阅读 2822·2019-08-23 18:26
阅读 1118·2019-08-23 18:20
阅读 2727·2019-08-23 16:43