Create a Highly Available Kubernetes Cluster Using Keepalived and HAproxy

A highly available Kubernetes cluster ensures your applications run without outages which is required for production. In this connection, there are plenty of ways for you to choose from to achieve high availability. For example, if your cluster is deployed on cloud (e.g. Google Cloud and AWS), you can create load balancers on these platforms directly. At the same time, Keepalived, HAproxy and NGINX are also possible alternatives for you to achieve load balancing.

In this article, I am going to use Keepalived and HAproxy for load balancing and achieve high availability. The steps are listed as below:

Prepare hosts.
Configure Keepalived and HAproxy.
Use KubeKey to set up a Kubernetes cluster.

Cluster Architecture

In my cluster, I will set three master nodes, three worker nodes, two nodes for load balancing and one virtual IP address. The virtual IP address in this example may also be called "a floating IP address". That means in the event of node failures, the IP address can be passed between nodes allowing for failover, thus achieving high availability.

architecture

Notice that in my cluster, I am not going to install Keepalived and HAproxy on any of the master nodes. Admittedly, you can do that and high availability can also be achieved. That said, I would like to try a different way by configuring two specific nodes for load balancing (You can add more nodes of this kind as needed). Only Keepalived and HAproxy will be installed on these two nodes, avoiding any potential conflicts with any Kubernetes components and services.

Host Information

Here is the detailed information of each node in my cluster for your reference:

IP Address	Host Name	Role	System
172.16.0.2	lb1	Keepalived & HAproxy	CentOS 7.5, 4 Cores, 4 G Memory, 20 G Disk
172.16.0.3	lb2	Keepalived & HAproxy	CentOS 7.5, 4 Cores, 4 G Memory, 20 G Disk
172.16.0.4	master1	master, etcd	CentOS 7.5, 8 Cores, 8 G Memory, 50 G Disk
172.16.0.5	master2	master, etcd	CentOS 7.5, 8 Cores, 8 G Memory, 50 G Disk
172.16.0.6	master3	master, etcd	CentOS 7.5, 8 Cores, 8 G Memory, 50 G Disk
172.16.0.7	worker1	worker	CentOS 7.5, 8 Cores, 8 G Memory, 50 G Disk
172.16.0.8	worker2	worker	CentOS 7.5, 8 Cores, 8 G Memory, 50 G Disk
172.16.0.9	worker3	worker	CentOS 7.5, 8 Cores, 8 G Memory, 50 G Disk
172.16.0.10		Virtual IP address

For more information about requirements for nodes, network, and dependencies, see one of my previous posts.

Configure Load Balancing

Keepalived provides a VRPP implementation and allows you to configure Linux machines for load balancing, preventing single points of failure. HAProxy, providing reliable, high performance load balancing, works perfectly with Keepalived.

As I said above, I will install both Keepalived and HAproxy on lb1 and lb2. The logic is very simple: if one of the node goes down, the virtual IP address (i.e. the floating IP address) will be automatically associated with another node so that the cluster is still functioning well, thus achieving high availability. If you want, you can add more nodes all with Keepalived and HAproxy installed for that purpose.

Run the following command to install Keepalived and HAproxy first.

yum install keepalived haproxy psmisc -y

HAproxy

The configuration of HAproxy is exactly the same on the two machines for load balancing. Run the following command to configure HAproxy.
```
vi /etc/haproxy/haproxy.cfg
```

Here is my configuration for your reference (Pay attention to the server field. Note that 6443 is the apiserver port):

global
    log /dev/log  local0 warning
    chroot      /var/lib/haproxy
    pidfile     /var/run/haproxy.pid
    maxconn     4000
    user        haproxy
    group       haproxy
    daemon

   stats socket /var/lib/haproxy/stats

defaults
  log global
  option  httplog
  option  dontlognull
        timeout connect 5000
        timeout client 50000
        timeout server 50000

frontend kube-apiserver
  bind *:6443
  mode tcp
  option tcplog
  default_backend kube-apiserver

backend kube-apiserver
    mode tcp
    option tcplog
    option tcp-check
    balance roundrobin
    default-server inter 10s downinter 5s rise 2 fall 2 slowstart 60s maxconn 250 maxqueue 256 weight 100
    server kube-apiserver-1 172.16.0.4:6443 check # Replace the IP address with your own.
    server kube-apiserver-2 172.16.0.5:6443 check # Replace the IP address with your own.
    server kube-apiserver-3 172.16.0.6:6443 check # Replace the IP address with your own.

Save the file and run the following command to restart HAproxy.
```
systemctl restart haproxy
```
Make it persist through reboots:
```
systemctl enable haproxy
```
Make sure you configure HAproxy on the other machine (lb2) as well.

Keepalived

Keepalived must be installed on both machines while the configuration of them is slightly different.

Run the following command to configure Keepalived.
```
vi /etc/keepalived/keepalived.conf
```

Here is my configuration (lb1) for your reference:

global_defs {
  notification_email {
  }
  router_id LVS_DEVEL
  vrrp_skip_check_adv_addr
  vrrp_garp_interval 0
  vrrp_gna_interval 0
}

vrrp_script chk_haproxy {
  script "killall -0 haproxy"
  interval 2
  weight 2
}

vrrp_instance haproxy-vip {
  state BACKUP
  priority 100
  interface eth0                       # Network card
  virtual_router_id 60
  advert_int 1
  authentication {
    auth_type PASS
    auth_pass 1111
  }
  unicast_src_ip 172.16.0.2      # The IP address of this machine
  unicast_peer {
    172.16.0.3                         # The IP address of peer machines
  }

  virtual_ipaddress {
    172.16.0.10/24                  # The VIP address
  }

  track_script {
    chk_haproxy
  }
}

Note

For the interface field, you must provide your own network card information. You can run ifconfig on your machine to get the value.
The IP address provided for unicast_src_ip is the IP address of your current machine. For other machines where HAproxy and Keepalived are also installed for load balancing, their IP address must be input for the field unicast_peer.

Save the file and run the following command to restart Keepalived.
```
systemctl restart keepalived
```
Make it persist through reboots:
```
systemctl enable haproxy
```
Make sure you configure Keepalived on the other machine (lb2) as well.

Verify HA

Before you start to create your Kubernetes cluster, make sure you have tested the high availability.

On the machine lb1, run the following command:

[root@lb1 ~]# ip a s
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: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    link/ether 52:54:9e:27:38:c8 brd ff:ff:ff:ff:ff:ff
    inet 172.16.0.2/24 brd 172.16.0.255 scope global noprefixroute dynamic eth0
       valid_lft 73334sec preferred_lft 73334sec
    inet 172.16.0.10/24 scope global secondary eth0 # The VIP address
       valid_lft forever preferred_lft forever
    inet6 fe80::510e:f96:98b2:af40/64 scope link noprefixroute
       valid_lft forever preferred_lft forever

As you can see above, the virtual IP address is successfully added. Simulate a failure on this node:
```
systemctl stop haproxy
```

Check the floating IP address again and you can see it disappear on lb1.

[root@lb1 ~]# ip a s
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: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    link/ether 52:54:9e:27:38:c8 brd ff:ff:ff:ff:ff:ff
    inet 172.16.0.2/24 brd 172.16.0.255 scope global noprefixroute dynamic eth0
       valid_lft 72802sec preferred_lft 72802sec
    inet6 fe80::510e:f96:98b2:af40/64 scope link noprefixroute
       valid_lft forever preferred_lft forever

Theoretically, the virtual IP will be failed over to the other machine (lb2) if the configuration is successful. On lb2, run the following command and here is the expected output:

[root@lb2 ~]# ip a s
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: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    link/ether 52:54:9e:3f:51:ba brd ff:ff:ff:ff:ff:ff
    inet 172.16.0.3/24 brd 172.16.0.255 scope global noprefixroute dynamic eth0
       valid_lft 72690sec preferred_lft 72690sec
    inet 172.16.0.10/24 scope global secondary eth0   # The VIP address
       valid_lft forever preferred_lft forever
    inet6 fe80::f67c:bd4f:d6d5:1d9b/64 scope link noprefixroute
       valid_lft forever preferred_lft forever

As you can see above, high availability is successfully configured.

Use KubeKey to Create a Kubernetes Cluster

KubeKey is an efficient and convenient tool to create a Kubernetes cluster. If you are not familiar with KubeKey, have a look at my previous articles about using KubeKey to create a three-node cluster and scale your cluster.

Download KubeKey from its GitHub Release Page or use the following command to download KubeKey version 1.0.1. You only need to download KubeKey to one of your machines (e.g. master1) that serves as the taskbox for installation.
```
curl -sfL https://get-kk.kubesphere.io | VERSION=v2.0.0 sh -
```
The above command downloads KubeKey and unzips the file. Your folder now contains a file called kk. Make it executable:
```
chmod +x kk
```
Create a configuration file to specify cluster information. The Kubernetes version I am going to install is v1.17.9.
```
./kk create config --with-kubernetes v1.20.4
```

A default file config-sample.yaml will be created. Edit the file and here is my configuration for your reference:

apiVersion: kubekey.kubesphere.io/v1alpha1
kind: Cluster
metadata:
  name: sample
spec:
  hosts:
  - {name: master1, address: 172.16.0.4, internalAddress: 172.16.0.4, user: root, password: Testing123}
  - {name: master2, address: 172.16.0.5, internalAddress: 172.16.0.5, user: root, password: Testing123}
  - {name: master3, address: 172.16.0.6, internalAddress: 172.16.0.6, user: root, password: Testing123}
  - {name: worker1, address: 172.16.0.7, internalAddress: 172.16.0.7, user: root, password: Testing123}
  - {name: worker2, address: 172.16.0.8, internalAddress: 172.16.0.8, user: root, password: Testing123}
  - {name: worker3, address: 172.16.0.9, internalAddress: 172.16.0.9, user: root, password: Testing123}
  roleGroups:
    etcd:
    - master1
    - master2
    - master3
    master:
    - master1
    - master2
    - master3
    worker:
    - worker1
    - worker2
    - worker3
  controlPlaneEndpoint:
    domain: lb.kubesphere.local
    address: 172.16.0.10   # The VIP address
    port: 6443
  kubernetes:
    version: v1.17.9
    imageRepo: kubesphere
    clusterName: cluster.local
  network:
    plugin: calico
    kubePodsCIDR: 10.233.64.0/18
    kubeServiceCIDR: 10.233.0.0/18
  registry:
    registryMirrors: []
    insecureRegistries: []
  addons: []

Note

Replace the value of controlPlaneEndpoint.address with your own VIP address.
For more information about different parameters in this configuration file, see one of my previous blogs.

Save the file and execute the following command to create your cluster:
```
./kk create cluster -f config-sample.yaml
```
You can see the output as below when the installation finishes.
```
Congratulations! Installation is successful.
```

Execute the following command to check the status of namespaces.

kubectl get pod --all-namespaces

NAMESPACE     NAME                                       READY   STATUS    RESTARTS   AGE
kube-system   calico-kube-controllers-59d85c5c84-l7zp5   1/1     Running   0          42s
kube-system   calico-node-5d6gb                          1/1     Running   0          21s
kube-system   calico-node-77bcj                          1/1     Running   0          42s
kube-system   calico-node-bdzfp                          1/1     Running   0          21s
kube-system   calico-node-ph756                          1/1     Running   0          22s
kube-system   calico-node-phz7d                          1/1     Running   0          22s
kube-system   calico-node-v7wnf                          1/1     Running   0          22s
kube-system   coredns-74d59cc5c6-gdkmz                   1/1     Running   0          53s
kube-system   coredns-74d59cc5c6-j2lhc                   1/1     Running   0          53s
kube-system   kube-apiserver-master1                     1/1     Running   0          48s
kube-system   kube-apiserver-master2                     1/1     Running   0          19s
kube-system   kube-apiserver-master3                     1/1     Running   0          19s
kube-system   kube-controller-manager-master1            1/1     Running   0          48s
kube-system   kube-controller-manager-master2            1/1     Running   0          19s
kube-system   kube-controller-manager-master3            1/1     Running   0          19s
kube-system   kube-proxy-29sfc                           1/1     Running   0          21s
kube-system   kube-proxy-drzsc                           1/1     Running   0          22s
kube-system   kube-proxy-lgwhd                           1/1     Running   0          22s
kube-system   kube-proxy-npq6t                           1/1     Running   0          21s
kube-system   kube-proxy-srlwx                           1/1     Running   0          22s
kube-system   kube-proxy-vdtbk                           1/1     Running   0          53s
kube-system   kube-scheduler-master1                     1/1     Running   0          48s
kube-system   kube-scheduler-master2                     1/1     Running   0          19s
kube-system   kube-scheduler-master3                     1/1     Running   0          20s
kube-system   nodelocaldns-2chnt                         1/1     Running   0          22s
kube-system   nodelocaldns-2wszl                         1/1     Running   0          22s
kube-system   nodelocaldns-2xqlc                         1/1     Running   0          21s
kube-system   nodelocaldns-92ksq                         1/1     Running   0          53s
kube-system   nodelocaldns-cktmd                         1/1     Running   0          22s
kube-system   nodelocaldns-skmlq                         1/1     Running   0          21s

Summary

Creating a highly available Kubernetes cluster is not just about business applications running without downtime. It is also about selecting the correct tools and using them to set up the cluster with high availability in the most graceful and efficient way. Why not try Keepalived, HAproxy and KubeKey? Perhaps they will give you the answer you have been seeking for so long.

Reference

KubeKey: A Lightweight Installer for Kubernetes and Cloud Native Addons

KubeKey GitHub Repository