CNLB-Lab : Create quick testbed to test k8s load-balancing implementations.

Pre-requisites

• Server/PC with Linux (> 16 cores, > 32Gb RAM)
• VirtualBox installed
• Vagrant installed

Supported LBs

• LoxiLB
• MetalLB

How to setup ??

git clone https://github.com/codeSnip12/cnlblab
cd cnlblab
vagrant up

The following setup would be created:

+--------------+        +---------------+
|    master    |        |   worker1     |
| 192.168.80.10|        | 192.168.80.101|
+------+-------+        +-------+-------+
       |                        |
       |                        |
       +-----------+------------+
                   |
                   |
             +-----+------+
             |   host     |
             |192.168.80.9|
             +------------+

How to setup and test LoxiLB ??

Login into master node

vagrant ssh master

Install LoxiLB and iperf pods

sudo kubectl apply -f /vagrant/manifests/kube-loxilb.yml 
sudo kubectl apply -f /vagrant/manifests/loxilb.yml
sudo kubectl apply -f /vagrant/manifests/iperf-service-loxilb.yml

Check everything is up and running

vagrant@master:~$ sudo kubectl get pods -A
NAMESPACE     NAME                                      READY   STATUS    RESTARTS   AGE
kube-system   local-path-provisioner-84db5d44d9-65kcg   1/1     Running   0          12m
kube-system   coredns-6799fbcd5-c5bs9                   1/1     Running   0          12m
kube-system   metrics-server-67c658944b-dtjkf           1/1     Running   0          12m
kube-system   kube-loxilb-745fb47486-nwmp8              1/1     Running   0          3m8s
kube-system   loxilb-lb-vkpm9                           1/1     Running   0          2m28s
kube-system   loxilb-lb-thpx8                           1/1     Running   0          2m28s

Run test

vagrant ssh host
$ iperf -c 192.168.80.20 -p 55001 -i1 -t10

Cleanup LoxiLB (if need to run MetalLB next) from master node

sudo kubectl delete -f /vagrant/manifests/iperf-service-loxilb.yml 
sudo kubectl delete -f /vagrant/manifests/loxilb.yml 
sudo kubectl delete -f /vagrant/manifests/kube-loxilb.yml 

How to setup and test MetalLB ??

Setup master node (MetalLB needs strictARP IPVS mode)

vagrant ssh master
$ sudo su 
$ echo 1 > /proc/sys/net/ipv4/conf/eth1/arp_ignore
$ echo 2 > /proc/sys/net/ipv4/conf/eth1/arp_announce

Setup worker node (MetalLB needs strictARP IPVS mode)

vagrant ssh worker1
$ sudo su 
$ echo 1 > /proc/sys/net/ipv4/conf/eth1/arp_ignore
$ echo 2 > /proc/sys/net/ipv4/conf/eth1/arp_announce

Install MetalLB and iperf pods from master node

sudo kubectl apply -f /vagrant/manifests/metallb-native.yaml
sudo kubectl apply -f /vagrant/manifests/metallb-addr-pool.yml
sudo kubectl apply -f /vagrant/manifests/iperf-service-metallb.yml 

Check everything is up and running

vagrant@master:~$ sudo kubectl get pods -A
NAMESPACE        NAME                                      READY   STATUS    RESTARTS   AGE
kube-system      local-path-provisioner-84db5d44d9-65kcg   1/1     Running   0          22m
kube-system      coredns-6799fbcd5-c5bs9                   1/1     Running   0          22m
kube-system      metrics-server-67c658944b-dtjkf           1/1     Running   0          22m
metallb-system   controller-786f9df989-pv2qv               1/1     Running   0          89s
metallb-system   speaker-lmr7r                             1/1     Running   0          89s
metallb-system   speaker-9nksg                             1/1     Running   0          89s

Run test

vagrant ssh host
$ iperf -c 192.168.80.20 -p 55001 -i1 -t10

How to move VIP assignment from one to another node

• For loxilb, check if VIP is assigned in lo interface of master/worker and delete pod loxilb-lb-xxxx running in that node.
• For metallb, could not find exact method. So, added static route for VIP (192.168.80.20 via 192.168.80.10 or 192.168.80.20 via 192.168.80.101) at host