mirror of
https://github.com/cloudnativelabs/kube-router.git
synced 2025-12-06 20:31:39 +01:00
3524a2a0ae
removed the requirement to edit kubelet manifest, as now kube-router gets node name from env variable
96 lines
5.9 KiB
Markdown
96 lines
5.9 KiB
Markdown
# Bootkube Integration
|
|
|
|
The following instructions and examples demonstrate how to create a new
|
|
[Bootkube](https://github.com/kubernetes-incubator/bootkube) provisioned
|
|
Kubernetes cluster using kube-router in place of kube-proxy and flannel.
|
|
|
|
## Asset Creation
|
|
|
|
Follow the
|
|
[documentation](https://github.com/kubernetes-incubator/bootkube#guides) for
|
|
your environment and setup arguments for `bootkube render`.
|
|
|
|
For example:
|
|
```
|
|
bootkube render --asset-dir=${PWD}/assets --api-servers=https://kube-api-dev.zbrbdl:443 --api-server-alt-names=DNS=kube-api-dev.zbrbdl --etcd-servers="http://127.0.0.1:2379"
|
|
Writing asset: /home/bzub/assets/manifests/kube-scheduler.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-scheduler-disruption.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-controller-manager-disruption.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-dns-deployment.yaml
|
|
Writing asset: /home/bzub/assets/manifests/pod-checkpointer.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-flannel.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-system-rbac-role-binding.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-controller-manager.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-apiserver.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-proxy.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-flannel-cfg.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-dns-svc.yaml
|
|
Writing asset: /home/bzub/assets/bootstrap-manifests/bootstrap-apiserver.yaml
|
|
Writing asset: /home/bzub/assets/bootstrap-manifests/bootstrap-controller-manager.yaml
|
|
Writing asset: /home/bzub/assets/bootstrap-manifests/bootstrap-scheduler.yaml
|
|
Writing asset: /home/bzub/assets/tls/ca.key
|
|
Writing asset: /home/bzub/assets/tls/ca.crt
|
|
Writing asset: /home/bzub/assets/tls/apiserver.key
|
|
Writing asset: /home/bzub/assets/tls/apiserver.crt
|
|
Writing asset: /home/bzub/assets/tls/service-account.key
|
|
Writing asset: /home/bzub/assets/tls/service-account.pub
|
|
Writing asset: /home/bzub/assets/tls/kubelet.key
|
|
Writing asset: /home/bzub/assets/tls/kubelet.crt
|
|
Writing asset: /home/bzub/assets/auth/kubeconfig
|
|
Writing asset: /home/bzub/assets/manifests/kube-apiserver-secret.yaml
|
|
Writing asset: /home/bzub/assets/manifests/kube-controller-manager-secret.yaml
|
|
```
|
|
|
|
## Kube-router Installation
|
|
|
|
Next move/delete the manifests for kube-proxy and flannel from
|
|
`assets/manifests` and replace them with the
|
|
[kube-router.yaml](/contrib/bootkube/kube-router.yaml) and
|
|
[kube-router-cfg.yaml](/contrib/bootkube/kube-router-cfg.yaml) files provided in
|
|
this repo.
|
|
```
|
|
rm assets/manifests/kube-flannel{,-cfg}.yaml assets/manifests/kube-proxy.yaml
|
|
curl -L https://raw.githubusercontent.com/cloudnativelabs/kube-router/master/contrib/bootkube/kube-router-cfg.yaml -o assets/manifests/kube-router-cfg.yaml
|
|
curl -L https://raw.githubusercontent.com/cloudnativelabs/kube-router/master/contrib/bootkube/kube-router.yaml -o assets/manifests/kube-router.yaml
|
|
```
|
|
|
|
## Additional steps when deploying on AWS
|
|
|
|
Since kube-router uses node routing rules to directly route pod-to-pod traffic to destination node, nodes will send and recieve IP traffic with source and destination IP's from pod CIDR. AWS by default prevents sending and recieving traffic from ip different from instance private ip. So we need to relax this restricition. Please run below commnad on each of the node in the cluster to send and recieve traffic from pod IP's.
|
|
|
|
```
|
|
aws ec2 modify-instance-attribute --instance-id <instance id>--no-source-dest-check
|
|
```
|
|
|
|
## Cluster Startup
|
|
|
|
Finally, proceed by following the Bootkube documentation, which generally
|
|
involves starting Kubelet and running `bootkube start` referring to your assets
|
|
directory on a new Kubernetes node.
|
|
|
|
After starting multiple master nodes, our example cluster looks like this:
|
|
```
|
|
$ kubectl -n kube-system get pods,services
|
|
NAME READY STATUS RESTARTS AGE IP NODE
|
|
po/kube-apiserver-gztjp 1/1 Running 0 15h 10.10.3.2 node2-dev.zbrbdl
|
|
po/kube-apiserver-h55t7 1/1 Running 0 15h 10.10.3.3 node3-dev.zbrbdl
|
|
po/kube-apiserver-qn5xm 1/1 Running 2 15h 10.10.3.1 node1-dev.zbrbdl
|
|
po/kube-controller-manager-3052101514-kp121 1/1 Running 1 15h 10.2.0.5 node1-dev.zbrbdl
|
|
po/kube-controller-manager-3052101514-n4q9p 1/1 Running 2 15h 10.2.0.6 node1-dev.zbrbdl
|
|
po/kube-dns-2431531914-pr9lg 3/3 Running 0 15h 10.2.0.3 node1-dev.zbrbdl
|
|
po/kube-router-ckdj1 1/1 Running 15 15h 10.10.3.3 node3-dev.zbrbdl
|
|
po/kube-router-dcgbr 1/1 Running 15 15h 10.10.3.1 node1-dev.zbrbdl
|
|
po/kube-router-n0vcn 1/1 Running 15 15h 10.10.3.2 node2-dev.zbrbdl
|
|
po/kube-scheduler-2172662190-g4q3w 1/1 Running 4 15h 10.2.0.2 node1-dev.zbrbdl
|
|
po/kube-scheduler-2172662190-hcq3t 1/1 Running 2 15h 10.2.0.4 node1-dev.zbrbdl
|
|
po/pod-checkpointer-jlfsv 1/1 Running 0 15h 10.10.3.1 node1-dev.zbrbdl
|
|
po/pod-checkpointer-jlfsv-node1-dev.zbrbdl 1/1 Running 0 15h 10.10.3.1 node1-dev.zbrbdl
|
|
po/pod-checkpointer-lhckt 1/1 Running 0 15h 10.10.3.3 node3-dev.zbrbdl
|
|
po/pod-checkpointer-lhckt-node3-dev.zbrbdl 1/1 Running 0 15h 10.10.3.3 node3-dev.zbrbdl
|
|
po/pod-checkpointer-tsbkh 1/1 Running 0 15h 10.10.3.2 node2-dev.zbrbdl
|
|
po/pod-checkpointer-tsbkh-node2-dev.zbrbdl 1/1 Running 0 15h 10.10.3.2 node2-dev.zbrbdl
|
|
|
|
NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
|
|
svc/kube-dns 10.3.0.10 <none> 53/UDP,53/TCP 15h k8s-app=kube-dns
|
|
```
|