跳转至

Mellanox 网卡 Offload 支持

Kube-OVN 在最终的数据平面使用 OVS 来完成流量转发,相关的流表匹配,隧道封装等功能为 CPU 密集型,在大流量下会消耗大量 CPU 资源并导致 延迟上升和吞吐量下降。Mellanox 的 Accelerated Switching And Packet Processing (ASAP²) 技术可以将 OVS 相关的操作卸载到硬件网卡内的 eSwitch 上执行。该技术可以在无需对 OVS 控制平面进行修改的情况下,缩短数据路径,避免对主机 CPU 资源的使用,大幅降低延迟并显著提升吞吐量。

前置条件

  • Mellanox CX5/CX6/CX7/BlueField 等支持 ASAP² 的硬件网卡。
  • CentOS 8 Stream 或上游 Linux 5.7 以上内核支持。
  • 由于当前网卡不支持 dp_hashhash 操作卸载,需关闭 OVN LB 功能。
  • 为了支持卸载模式,网卡不能做 bond。

配置 SR-IOV 和 Device Plugin

Mellanox 网卡支持两种配置 offload 的方式,一种手动配置网卡 SR-IOV 和 Device Plugin,另一种通过 sriov-network-operator 进行自动配置。

手动配置 SR-IOV 和 Device Plugin

查询网卡的设备 ID,下面的例子中为 42:00.0

# lspci -nn | grep ConnectX-5
42:00.0 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5] [15b3:1017]

根据设备 ID 找到对应网卡:

# ls -l /sys/class/net/ | grep 42:00.0
lrwxrwxrwx. 1 root root 0 Jul 22 23:16 p4p1 -> ../../devices/pci0000:40/0000:40:02.0/0000:42:00.0/net/p4p1

检查可用 VF 数量:

# cat /sys/class/net/p4p1/device/sriov_totalvfs
8

创建 VF,总数不要超过上面查询出的数量:

# echo '4' > /sys/class/net/p4p1/device/sriov_numvfs
# ip link show p4p1
10: p4p1: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc mq state DOWN mode DEFAULT group default qlen 1000
    link/ether b8:59:9f:c1:ec:12 brd ff:ff:ff:ff:ff:ff
    vf 0 MAC 00:00:00:00:00:00, spoof checking off, link-state auto, trust off, query_rss off
    vf 1 MAC 00:00:00:00:00:00, spoof checking off, link-state auto, trust off, query_rss off
    vf 2 MAC 00:00:00:00:00:00, spoof checking off, link-state auto, trust off, query_rss off
    vf 3 MAC 00:00:00:00:00:00, spoof checking off, link-state auto, trust off, query_rss off
# ip link set p4p1 up

找到上述 VF 对应的设备 ID:

# lspci -nn | grep ConnectX-5
42:00.0 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5] [15b3:1017]
42:00.1 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5] [15b3:1017]
42:00.2 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
42:00.3 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
42:00.4 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
42:00.5 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]

将 VF 从驱动中解绑:

echo 0000:42:00.2 > /sys/bus/pci/drivers/mlx5_core/unbind
echo 0000:42:00.3 > /sys/bus/pci/drivers/mlx5_core/unbind
echo 0000:42:00.4 > /sys/bus/pci/drivers/mlx5_core/unbind
echo 0000:42:00.5 > /sys/bus/pci/drivers/mlx5_core/unbind

开启 eSwitch 模式,并设置硬件卸载:

devlink dev eswitch set pci/0000:42:00.0 mode switchdev
ethtool -K enp66s0f0 hw-tc-offload on

重新绑定驱动,完成 VF 设置:

echo 0000:42:00.2 > /sys/bus/pci/drivers/mlx5_core/bind
echo 0000:42:00.3 > /sys/bus/pci/drivers/mlx5_core/bind
echo 0000:42:00.4 > /sys/bus/pci/drivers/mlx5_core/bind
echo 0000:42:00.5 > /sys/bus/pci/drivers/mlx5_core/bind

NetworkManager 的一些行为可能会导致驱动异常,如果卸载出现问题建议关闭 NetworkManager 再进行尝试:

systemctl stop NetworkManager
systemctl disable NetworkManager

由于每个机器的 VF 数量优先,每个使用加速的 Pod 会占用 VF 资源,我们需要使用 SR-IOV Device Plugin 管理相应资源,使得调度器知道如何根据 资源进行调度。

创建 SR-IOV 相关 Configmap:

apiVersion: v1
kind: ConfigMap
metadata:
  name: sriovdp-config
  namespace: kube-system
data:
  config.json: |
    {
      "resourceList": [{
          "resourcePrefix": "mellanox.com",
          "resourceName": "cx5_sriov_switchdev",
          "selectors": {
                  "vendors": ["15b3"],
                  "devices": ["1018"],
                  "drivers": ["mlx5_core"]
              }
      }
      ]
    }

参考 SR-IOV 文档进行部署:

kubectl apply -f https://raw.githubusercontent.com/intel/sriov-network-device-plugin/master/deployments/k8s-v1.16/sriovdp-daemonset.yaml

检查 SR-IOV 资源是否已经注册到 Kubernetes Node 中:

kubectl describe node kube-ovn-01  | grep mellanox

mellanox.com/cx5_sriov_switchdev:  4
mellanox.com/cx5_sriov_switchdev:  4
mellanox.com/cx5_sriov_switchdev  0           0

使用 sriov-network-operator 配置 SR-IOV 和 Device Plugin

安装 node-feature-discovery 自动检测硬件的功能和系统配置:

kubectl apply -k https://github.com/kubernetes-sigs/node-feature-discovery/deployment/overlays/default?ref=v0.11.3

或者通过下面的命令,手动给有 offload 能力的网卡增加 annotation:

kubectl label nodes [offloadNicNode] feature.node.kubernetes.io/network-sriov.capable=true

克隆代码仓库并安装 Operator:

git clone --depth=1 https://github.com/kubeovn/sriov-network-operator.git
kubectl apply -k sriov-network-operator/deploy

检查 Operator 组件是否工作正常:

# kubectl get -n kube-system all | grep sriov
NAME                                          READY   STATUS    RESTARTS   AGE
pod/sriov-network-config-daemon-bf9nt         1/1     Running   0          8s
pod/sriov-network-operator-54d7545f65-296gb   1/1     Running   0          10s

NAME                                         DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR                                                 AGE
daemonset.apps/sriov-network-config-daemon   1         1         1       1            1           beta.kubernetes.io/os=linux,feature.node.kubernetes.io/network-sriov.capable=true   8s

NAME                                     READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/sriov-network-operator   1/1     1            1           10s

NAME                                                DESIRED   CURRENT   READY   AGE
replicaset.apps/sriov-network-operator-54d7545f65   1         1         1       10s

检查 SriovNetworkNodeState,下面以 node1 节点为例,该节点上有两个 Mellanox 网卡:

# kubectl get sriovnetworknodestates.sriovnetwork.openshift.io -n kube-system node1 -o yaml
apiVersion: sriovnetwork.openshift.io/v1
kind: SriovNetworkNodeState
spec: ...
status:
  interfaces:
  - deviceID: "1017"
    driver: mlx5_core
    mtu: 1500
    pciAddress: "0000:5f:00.0"
    totalvfs: 8
    vendor: "15b3"
    linkSeed: 25000Mb/s
    linkType: ETH
    mac: 08:c0:eb:f4:85:bb
    name: ens41f0np0
  - deviceID: "1017"
    driver: mlx5_core
    mtu: 1500
    pciAddress: "0000:5f:00.1"
    totalvfs: 8
    vendor: "15b3"
    linkSeed: 25000Mb/s
    linkType: ETH
    mac: 08:c0:eb:f4:85:bb
    name: ens41f1np1

创建 SriovNetworkNodePolicy 资源,并通过 nicSelector 选择要管理的网卡:

apiVersion: sriovnetwork.openshift.io/v1
kind: SriovNetworkNodePolicy
metadata:
  name: policy
  namespace: kube-system
spec:
  nodeSelector:
    feature.node.kubernetes.io/network-sriov.capable: "true"
  eSwitchMode: switchdev
  numVfs: 3
  nicSelector:
    pfNames:
    - ens41f0np0
    - ens41f1np1
  resourceName: cx_sriov_switchdev

再次检查 SriovNetworkNodeStatestatus 字段:

# kubectl get sriovnetworknodestates.sriovnetwork.openshift.io -n kube-system node1 -o yaml

...
spec:
  interfaces:
  - eSwitchMode: switchdev
    name: ens41f0np0
    numVfs: 3
    pciAddress: 0000:5f:00.0
    vfGroups:
    - policyName: policy
      vfRange: 0-2
      resourceName: cx_sriov_switchdev
  - eSwitchMode: switchdev
    name: ens41f1np1
    numVfs: 3
    pciAddress: 0000:5f:00.1
    vfGroups:
    - policyName: policy
      vfRange: 0-2
      resourceName: cx_sriov_switchdev
status:
  interfaces
  - Vfs:
    - deviceID: 1018
      driver: mlx5_core
      pciAddress: 0000:5f:00.2
      vendor: "15b3"
    - deviceID: 1018
      driver: mlx5_core
      pciAddress: 0000:5f:00.3
      vendor: "15b3"
    - deviceID: 1018
      driver: mlx5_core
      pciAddress: 0000:5f:00.4
      vendor: "15b3"
    deviceID: "1017"
    driver: mlx5_core
    linkSeed: 25000Mb/s
    linkType: ETH
    mac: 08:c0:eb:f4:85:ab
    mtu: 1500
    name: ens41f0np0
    numVfs: 3
    pciAddress: 0000:5f:00.0
    totalvfs: 3
    vendor: "15b3"
  - Vfs:
    - deviceID: 1018
      driver: mlx5_core
      pciAddress: 0000:5f:00.5
      vendor: "15b3"
    - deviceID: 1018
      driver: mlx5_core
      pciAddress: 0000:5f:00.6
      vendor: "15b3"
    - deviceID: 1018
      driver: mlx5_core
      pciAddress: 0000:5f:00.7
      vendor: "15b3"
    deviceID: "1017"
    driver: mlx5_core
    linkSeed: 25000Mb/s
    linkType: ETH
    mac: 08:c0:eb:f4:85:bb
    mtu: 1500
    name: ens41f1np1
    numVfs: 3
    pciAddress: 0000:5f:00.1
    totalvfs: 3
    vendor: "15b3"

检查 VF 的状态:

# lspci -nn | grep ConnectX
5f:00.0 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5] [15b3:1017]
5f:00.1 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5] [15b3:1017]
5f:00.2 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
5f:00.3 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
5f:00.4 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
5f:00.5 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
5f:00.6 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]
5f:00.7 Ethernet controller [0200]: Mellanox Technologies MT27800 Family [ConnectX-5 Virtual Function] [15b3:1018]

检查 PF 工作模式:

# cat /sys/class/net/ens41f0np0/compat/devlink/mode
switchdev

安装 Multus-CNI

SR-IOV Device Plugin 调度时获得的设备 ID 需要通过 Multus-CNI 传递给 Kube-OVN,因此需要配置 Multus-CNI 配合完成相关任务。

参考 Multus-CNI 文档进行部署:

kubectl apply -f https://raw.githubusercontent.com/k8snetworkplumbingwg/multus-cni/master/deployments/multus-daemonset.yml

创建 NetworkAttachmentDefinition

apiVersion: "k8s.cni.cncf.io/v1"
kind: NetworkAttachmentDefinition
metadata:
  name: default
  namespace: default
  annotations:
    k8s.v1.cni.cncf.io/resourceName: mellanox.com/cx5_sriov_switchdev
spec:
  config: '{
    "cniVersion": "0.3.1",
    "name": "kube-ovn",
    "plugins":[
        {
            "type":"kube-ovn",
            "server_socket":"/run/openvswitch/kube-ovn-daemon.sock",
            "provider": "default.default.ovn"
        },
        {
            "type":"portmap",
            "capabilities":{
                "portMappings":true
            }
        }
    ]
}'
  • provider: 格式为当前 NetworkAttachmentDefinition 的 {name}.{namespace}.ovn。

Kube-OVN 中开启卸载模式

下载安装脚本:

wget https://raw.githubusercontent.com/alauda/kube-ovn/release-1.11/dist/images/install.sh

修改相关参数,IFACE 需要为物理网卡名,该网卡需要有可路由 IP:

ENABLE_MIRROR=${ENABLE_MIRROR:-false}
HW_OFFLOAD=${HW_OFFLOAD:-true}
ENABLE_LB=${ENABLE_LB:-false}
IFACE="ensp01"

安装 Kube-OVN:

bash install.sh

创建使用 VF 网卡的 Pod

可以使用如下 yaml 格式创建使用 VF 进行网络卸载加速的 Pod:

apiVersion: v1
kind: Pod
metadata:
  name: nginx
  annotations:
    v1.multus-cni.io/default-network: default/default
spec:
  containers:
  - name: nginx
    image: docker.io/library/nginx:alpine
    resources:
      requests:
        mellanox.com/cx5_sriov_switchdev: '1'
      limits:
        mellanox.com/cx5_sriov_switchdev: '1'
  • v1.multus-cni.io/default-network: 为上一步骤中 NetworkAttachmentDefinition 的 {namespace}/{name}。

可通过在 Pod 运行节点的 ovs-ovn 容器中运行下面的命令观察卸载是否成功:

# ovs-appctl dpctl/dump-flows -m type=offloaded
ufid:91cc45de-e7e9-4935-8f82-1890430b0f66, skb_priority(0/0),skb_mark(0/0),ct_state(0/0x23),ct_zone(0/0),ct_mark(0/0),ct_label(0/0x1),recirc_id(0),dp_hash(0/0),in_port(5b45c61b307e_h),packet_type(ns=0/0,id=0/0),eth(src=00:00:00:c5:6d:4e,dst=00:00:00:e7:16:ce),eth_type(0x0800),ipv4(src=0.0.0.0/0.0.0.0,dst=0.0.0.0/0.0.0.0,proto=0/0,tos=0/0,ttl=0/0,frag=no), packets:941539, bytes:62142230, used:0.260s, offloaded:yes, dp:tc, actions:54235e5753b8_h
ufid:e00768d7-e652-4d79-8182-3291d852b791, skb_priority(0/0),skb_mark(0/0),ct_state(0/0x23),ct_zone(0/0),ct_mark(0/0),ct_label(0/0x1),recirc_id(0),dp_hash(0/0),in_port(54235e5753b8_h),packet_type(ns=0/0,id=0/0),eth(src=00:00:00:e7:16:ce,dst=00:00:00:c5:6d:4e),eth_type(0x0800),ipv4(src=0.0.0.0/0.0.0.0,dst=0.0.0.0/0.0.0.0,proto=0/0,tos=0/0,ttl=0/0,frag=no), packets:82386659, bytes:115944854173, used:0.260s, offloaded:yes, dp:tc, actions:5b45c61b307e_h

如果有 offloaded:yes, dp:tc 内容证明卸载成功。

微信群 Slack Twitter Support


最后更新: 2023年5月9日
创建日期: 2022年5月24日

评论

回到页面顶部