https://github.com/aledbf/kube-haproxy-router

https://github.com/aledbf/kube-haproxy-router

Last synced: 3 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/aledbf/kube-haproxy-router
• Owner: aledbf
• Created: 2015-08-03T13:54:38.000Z (over 9 years ago)
• Default Branch: master
• Last Pushed: 2015-11-01T21:09:56.000Z (about 9 years ago)
• Last Synced: 2024-04-14T16:09:07.223Z (9 months ago)
• Language: Go
• Size: 2.88 MB
• Stars: 6
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
## DEPRECATED 

# The contents of this repository is in this  [kubernetes/contrib](https://github.com/kubernetes/contrib/tree/master/service-loadbalancer) repository

# kube-haproxy-router

Replace nginx to avoid kube-proxy to reach a pod.

`Internet-> Nginx -> iptables (VIP) -> kube-proxy -> go LB -> flannel -> docker bridge -> pod/s`

with

`Internet -> kube-haproxy -> flannel -> docker bridge -> pod/s`

**Changes:**

  - remove nginx

  - do not use service VIP to reach a pod

  - use the name of the application (hostname) to route traffic

  - use haproxy checks

This borrow ideas from #12111 and #11679 in `github.com/GoogleCloudPlatform/kubernetes`

**Content type**

Some applications, like a REST API always return json content. In this cases it makes sense to also return error (404,408,502,503 and 504) in the same format.

Any `kubernetes` service with the label `content=json` will return a json response in case of an error.

```

docker run \

  --name kube-haproxy \

  --rm \

  -p 80:80 \

  -p 443:443 \

  -p 1936:1936 \

  -p 2222:2222 \

  -p 8081:8081 \

  -v /dev/log:/dev/log \

  aledbf/kube-haproxy-router:v0.0.1 \

  /kube-haproxy \

  --master http://$(etcdctl get /deis/scheduler/k8s/master):8080 \

  --domain=$(etcdctl get /deis/platform/domain) \

  --nodes=$(fleetctl list-machines -fields=ip -no-legend | xargs | sed -e 's/ /,/g')

```