k8s whereabouts 使用

Whereabouts简介

whereabouts是一个cluster-wide（集群级别）的IPAM插件，非常适合用在NetworkAttachment的场景。之前我们描述过k8s为分配地址使用的是ipam，常见的ipam类型为host-local，calico-ipam。whereabouts是一款用于替换host-local的ipam。记录本地或者k8s已经分配的pool和ip地址。通常以 NAD 为地址段设定的模式，一个 nad 中描述该地址段的 ip 范围，网关等信息。一般和cni bridge，ipvlan，macvlan 等相结合使用，负责分配和管理 IP 的功能。

一、部署whereabouts

下载安装包

1：克隆代码
[root@node1 ~]# git clone https://github.com/k8snetworkplumbingwg/whereabouts
Cloning into 'whereabouts'...
remote: Enumerating objects: 23528, done.
remote: Counting objects: 100% (3556/3556), done.
remote: Compressing objects: 100% (1875/1875), done.
remote: Total 23528 (delta 1811), reused 2957 (delta 1580), pack-reused 19972
Receiving objects: 100% (23528/23528), 36.48 MiB | 9.56 MiB/s, done.
Resolving deltas: 100% (11207/11207), done.

#################################
2；查看文件
[root@node1 ~]# cd whereabouts/doc/crds/
[root@node1 crds]# ll
total 12
-rw-r--r-- 1 root root 2662 Nov 29 14:27 daemonset-install.yaml
-rw-r--r-- 1 root root 2566 Nov 29 14:27 whereabouts.cni.cncf.io_ippools.yaml
-rw-r--r-- 1 root root 2039 Nov 29 14:27 whereabouts.cni.cncf.io_overlappingrangeipreservations.yaml

##########################################
3：执行以下命令开始部署whereabouts
[root@node1 crds]# kubectl apply -f daemonset-install.yaml
serviceaccount/whereabouts created
clusterrolebinding.rbac.authorization.k8s.io/whereabouts created
clusterrole.rbac.authorization.k8s.io/whereabouts-cni created
daemonset.apps/whereabouts created
[root@node1 crds]# kubectl apply -f whereabouts.cni.cncf.io_ippools.yaml
customresourcedefinition.apiextensions.k8s.io/ippools.whereabouts.cni.cncf.io created
[root@node1 crds]# kubectl apply -f whereabouts.cni.cncf.io_overlappingrangeipreservations.yaml
customresourcedefinition.apiextensions.k8s.io/overlappingrangeipreservations.whereabouts.cni.cncf.io created
查看生成的crd

1：查看生成的crd
[root@node1 crds]# kubectl get crd | grep whereabouts
ippools.whereabouts.cni.cncf.io 2023-11-29T06:50:47Z
overlappingrangeipreservations.whereabouts.cni.cncf.io 2023-11-29T06:50:52Z
[root@node1 crds]#

################解释：
overlappingrangeipreservations 所有分配的 IP
Ippool 已经分配的ip的地址池
查看服务

1；查看whereabouts服务运行状态，以daemonset形式在每个节点运行
[root@node1 ~]# kubectl get po -A -o wide | grep whereabouts
kube-system whereabouts-4sjpf 1/1 Running 0 19m 192.168.5.27 node3 <none> <none>
kube-system whereabouts-rw5d5 1/1 Running 0 19m 192.168.5.126 node2 <none> <none>
kube-system whereabouts-v8nxd 1/1 Running 0 19m 192.168.5.79 node1 <none>

##########################################
2：会在每个节点的/opt/cni/bin 目录下生成一个二进制文件
[root@node1 ~]# cd /opt/cni/bin/
[root@node1 bin]# ll | grep where
-rwxr-xr-x 1 root root 45792495 Nov 29 15:03 whereabouts
[root@node1 bin]#

##########解释
daemonset服务负责维护地址池
whereabouts插件负责申请地址

二、实验

以下测试皆是基于生产环境来配置。

实验环境基于kube-ovn + multus + ipvlan给容器设置双网卡，ipvlan使用whereabouts来分配置IP

网络规划

在我的环境中，ipvlan使用的是两块网卡做的bond，每个子接口走不同的业务，本次测试使用的是ipvl1720
[root@node1 ~]# ip a | grep bond_ipvl
2: enp35s0f0: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 1500 qdisc mq master bond_ipvl state UP group default qlen 1000
9: enp97s0f1: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 1500 qdisc mq master bond_ipvl state UP group default qlen 1000
77: bond_ipvl: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
78: bond_ipvl.1520@bond_ipvl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
79: bond_ipvl.1511@bond_ipvl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
80: bond_ipvl.1510@bond_ipvl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
81: bond_ipvl.1521@bond_ipvl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
82: bond_ipvl.1522@bond_ipvl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
207: bond_ipvl.1720@bond_ipvl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
创建NAD

[root@node1 ~]# cat ipvlan1720.yaml
apiVersion: k8s.cni.cncf.io/v1
kind: NetworkAttachmentDefinition
metadata:
name: ipvlan1720
namespace: devops
spec:
config: '{ "cniVersion": "0.3.1",
"LogFile": "/var/log/multus.log",
"LogLevel": "debug",
"name": "ipvlan1720",
"type": "ipvlan",
"master": "bond_ipvl.1720", ##使用的接口
"mtu": 1500,
"ipam": { "type": "whereabouts", ###指定ipam类型
"datastore": "kubernetes",
"range": "10.11.88.96/28", ###cidr
"range_start": "10.11.88.98", ###起始地址
"range_end": "10.11.88.98", ###终止地址
"gateway": "10.11.8.97", ###网关
"log_file" : "/tmp/whereabouts.log",
"log_level" : "debug"
}
}'

[root@node1 ~]# kubectl apply -f ipvl1720.yaml
networkattachmentdefinition.k8s.cni.cncf.io/ipvlan1720 created
[root@node1 ~]#
[root@node1 ~]#
[root@node1 ~]# kubectl get network-attachment-definition -n devops
NAME AGE
ipvlan1720 78s
[root@node1 ~]#
启动pod

1；pod yaml内容如下
apiVersion: v1
kind: Pod
metadata:
name: ipvl
namespace: devops
labels:
app: nginx
annotations:
k8s.v1.cni.cncf.io/networks: ipvlan1720 ###指定使用的NAD
spec:
containers:
- name: nginx
image: registry-1.ict-mec.net:18443/nginx:latest
imagePullPolicy: IfNotPresent
ports:
- name: nginx-port
containerPort: 80

####################################
2：查看pod,172.10.4.148为kube-ovn网络的ip地址
[root@node1 ~]# kubectl get po -n devops
NAME READY STATUS RESTARTS AGE
ipvl 1/1 Running 0 3s
[root@node1 ~]# kubectl get po -n devops -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
ipvl 1/1 Running 0 6s 172.10.4.148 node2 <none> <none>
[root@node1 ~]#
查看pod获取的地址

1:describe 查看pod信息
[root@node1 ~]# kubectl describe po ipvl -n devops
Name: ipvl
Namespace: devops
Priority: 0
Node: cmu52/23.1.1.36
Start Time: Thu, 30 Nov 2023 15:45:33 +0800
Labels: app=nginx
Annotations: io.kubernetes.pod.sandbox.uid: aa3576c85e9b699a8bbe5360edd5d84cccffdfc2e1c017b6b6e52c77c0308dcd
k8s.v1.cni.cncf.io/network-status:
[{
"name": "kube-ovn",
"ips": [
"172.10.4.148",
"fd00:10:16::494"
],
"default": true,
"dns": {}
},{
"name": "devops/ipvlan1720",
"interface": "net1",
"ips": [
"10.11.88.98"
],
"mac": "40:a6:b7:37:29:3c",
"dns": {}
}]
k8s.v1.cni.cncf.io/networks: ipvlan1720 ###使用的NAD
k8s.v1.cni.cncf.io/networks-status:
[{
"name": "kube-ovn",
"ips": [
"172.10.4.148",
"fd00:10:16::494"
],
"default": true,
"dns": {}
},{
"name": "devops/ipvlan1720",
"interface": "net1", ####pod内部的网卡
"ips": [
"10.11.88.98" ####可以看到获取的地址
],
"mac": "40:a6:b7:37:29:3c",
"dns": {}
}]

###############################
2：进入pod的宿主机的网络ns 中查看
[root@node2 ~]# crictl ps| grep nginx
38a9298403eda 12766a6745eea About a minute ago Running nginx 0 aa3576c85e9b6 ipvl
[root@node2 ~]# crictl inspect 38a9298403eda| grep -i pid
"pid": 505586,
"pid": 1
"type": "pid"
[root@node2 ~]# nsenter -t 505586 -n bash
[root@node2 ~]# ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
inet6 ::1/128 scope host
valid_lft forever preferred_lft forever
2: ip_vti0@NONE: <NOARP> mtu 1480 qdisc noop state DOWN group default qlen 1000
link/ipip 0.0.0.0 brd 0.0.0.0
3: net1@if248: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN group default
link/ether 40:a6:b7:37:29:3c brd ff:ff:ff:ff:ff:ff
inet 10.11.88.98/28 brd 10.11.88.111 scope global net1
valid_lft forever preferred_lft forever
inet6 fe80::40a6:b700:137:293c/64 scope link
valid_lft forever preferred_lft forever
259: eth0@if260: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 qdisc noqueue state UP group default
link/ether 00:00:00:20:4f:93 brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 172.10.4.148/16 brd 172.10.255.255 scope global eth0
valid_lft forever preferred_lft forever
inet6 fd00:10:16::494/64 scope global
valid_lft forever preferred_lft forever
inet6 fe80::200:ff:fe20:4f93/64 scope link
valid_lft forever preferred_lft forever
[root@node2 ~]# ip r
default via 172.10.0.1 dev eth0
10.11.88.96/28 dev net1 proto kernel scope link src 10.11.88.98
172.10.0.0/16 dev eth0 proto kernel scope link src 172.10.4.148
查看分配的ip地址

1：查看已经分配的地址
[root@node1 ~]# kubectl get overlappingrangeipreservations.whereabouts.cni.cncf.io -n kube-system
NAME AGE
10.11.88.98 79m

################
2：查看已经分配的地址池，因为我在上面的NAD中定义的网络为10.11.88.96/28，所以pool的名称规则是网络+掩码位数
[root@node1 ~]# kubectl get ippools.whereabouts.cni.cncf.io -n kube-system
NAME AGE
10.11.88.96-28 80m

####################################
3:查看分配的ip地址信息
[root@node1 ~]# kubectl describe overlappingrangeipreservations.whereabouts.cni.cncf.io 10.11.88.98 -n kube-system
Name: 10.11.88.98
Namespace: kube-system
Labels: <none>
Annotations: <none>
API Version: whereabouts.cni.cncf.io/v1alpha1
Kind: OverlappingRangeIPReservation
Metadata:
Creation Timestamp: 2023-11-30T07:45:34Z
Generation: 1
Managed Fields:
API Version: whereabouts.cni.cncf.io/v1alpha1
Fields Type: FieldsV1
fieldsV1:
f:spec:
.:
f:containerid:
f:podref:
Manager: whereabouts
Operation: Update
Time: 2023-11-30T07:45:34Z
Resource Version: 25214006
UID: d1565403-02e1-45ff-b09f-a26710d7bb7d
Spec:
Containerid: aa3576c85e9b699a8bbe5360edd5d84cccffdfc2e1c017b6b6e52c77c0308dcd ####可以看到绑定的容器id
Podref: devops/ipvl
Events: <none>
[root@node1 ~]#

##############################
4：查看地址池的信息
[root@node1 ~]# kubectl describe ippools.whereabouts.cni.cncf.io 10.11.88.96-28 -n kube-system
Name: 10.11.88.96-28
Namespace: kube-system
Labels: <none>
Annotations: <none>
API Version: whereabouts.cni.cncf.io/v1alpha1
Kind: IPPool
Metadata:
Creation Timestamp: 2023-11-30T07:44:44Z
Generation: 4
Managed Fields:
API Version: whereabouts.cni.cncf.io/v1alpha1
Fields Type: FieldsV1
fieldsV1:
f:spec:
.:
f:allocations:
.:
f:2:
.:
f:id:
f:podref:
f:range:
Manager: whereabouts
Operation: Update
Time: 2023-11-30T07:45:34Z
Resource Version: 25214005
UID: 10544b9b-9ad3-43a9-b747-d46293a54ebb
Spec:
Allocations:
2:
Id: aa3576c85e9b699a8bbe5360edd5d84cccffdfc2e1c017b6b6e52c77c0308dcd
Podref: devops/ipvl ###pod的名称
Range: 10.11.88.96/28
Events: <none>
[root@node1 ~]#
pod删除

1：测试多启动一个pod，发现在ippool 中会自动把发分配的ip更新进去
[root@node1 ~]# kubectl get ippool 10.11.88.96-28 -o yaml -n kube-system
apiVersion: whereabouts.cni.cncf.io/v1alpha1
kind: IPPool
metadata:
creationTimestamp: "2023-11-30T09:38:48Z"
generation: 3
name: 10.11.88.96-28
namespace: kube-system
resourceVersion: "25262851"
uid: d00b0b1c-f848-461b-9e98-e4512307a485
spec:
allocations:
"2":
id: cc4d323c718f08effc22ff05f4a7727a9b648e5bd4849ab0143c5439337209d1
podref: devops/ipvl
"3":
id: 7a8bbfed7f98586779ac089358102078bca0f1946fe131c6abd8bd66e685688a
podref: devops/ipvl2
range: 10.11.88.96/28

########################################
2：删除两个pod，查看ippool的变化
[root@node1 ~]# kubectl delete po ipvl -n devops
pod "ipvl" deleted
[root@cmu51 ~]# kubectl delete po ipvl2 -n devops
pod "ipvl2" deleted

如下ippool中已经将分配到ip地址进行删除。
[root@node1 ~]# kubectl get ippool 10.11.88.96-28 -o yaml -n kube-system
apiVersion: whereabouts.cni.cncf.io/v1alpha1
kind: IPPool
metadata:
creationTimestamp: "2023-11-30T09:38:48Z"
generation: 5
name: 10.11.88.96-28
namespace: kube-system
resourceVersion: "25263222"
uid: d00b0b1c-f848-461b-9e98-e4512307a485
spec:
allocations: {}
range: 10.11.88.96/28

三、原理总结

1：创建一个NAD，内部定义一个IP range

2：创建pod，调度到宿主机时，本机的二进制文件whereabouts读取ip range

3：获取ippool，如果不存在就创建一个新的ippool,比如NAD中定义的IP range为10.11.88.96/28，ippool的名称为10.11.88.96-28

4：获取到 ippool 之后，根据NAD.ipam 地址段里的 IP range 里分配一个可用 ip,将 ip 更新到 ippool 里。

5：创建 OverlappingRangeIPReservation，名称以IP地址命名。

6：cni将结果返回给cri