Add example backend service

[philhug]

It would be nice to also have an example service file. I seem to have problems getting it to work:

kind: Service
apiVersion: v1
metadata:
  name: test-vulcand-proxy
  annotations:
    vulcand.io/routed: "true"
    vulcand.io/route: Host("example.com")
spec:
  ports:
  - name: http
    port: 80
    targetPort: 80
    protocol: TCP
  selector:
    name: test
  type: ClusterIP
  ClusterIP: None

time="2015-11-20T14:39:07Z" level=info msg="Adding service" serviceID=default-test-vulcand-proxy
time="2015-11-20T14:39:07Z" level=warning msg="Unable to find service for endpoint" name=test-vulcand-proxy namespace=default
time="2015-11-20T14:39:07Z" level=warning msg="Could not get endpoints for service" error= serviceID=default-test-vulcand-proxy

[ekozan]

you have forget to add your selector

{
    "kind": "Service",
    "apiVersion": "v1",
    "metadata": {
        "name": "test-vulcand-proxy",
        "vulcand.io/routed": "true",
        "vulcand.io/route": "Host('example.com')"
    },
    "spec": {
        "selector": {
            "app": "test-vulcand-proxy"
        },
        "ports": [
            {
                "protocol": "TCP",
                "port": 80,
                "targetPort": 80
            }
        ]
    }
}

and be sure you have somme pods running

[philhug]

the selector is there. the formatting was just messed up:
selector:
name: test

[ghost]

Hey guys,

Here are 2 great utils that woudl be great for kubernetes-vulcand-router.

Vulcan-Config-Builder

Also, for the vulcand config itself, Financial Times released that some util that could be also interesting for this repo. (Large scale registration): https://github.com/Financial-Times/vulcan-config-builder

Romulus

And you have also https://github.com/timelinelabs/romulus/ that allows to automatically register a kubernetes services in vulcan proxy .

Cheers,
Luc Michalski

[emmanuel]

@lucmichalski thanks for bring romulus to my attention. The FT thing is interesting as well.

😢 I was not aware of romulus when developing this. If I had seen it, I would have used romulus and not have written this. I'm deprecating this code and will not be making any further improvements to it. Please don't use this (kubernetes-vulcand-router), use romulus instead (which appears to do everything this code does, and more).

[ghost]

Emmanuel,

At the opposite, I think that you provided an awesome piece of the puzzle in order to make of Vulcan a serious leading proxy for smart micro-services. So, I think that your repo is 30% of the whole as it own as it sticks to kubernetes as a boostrap template.

Maybe a more advanced tool would help Vulcan for sure. As I think that circuitbreakers, middlewares make me think about an web UI, link tensorflow (ex; http://images.csmonitor.com/csm/2015/11/946808_1_1109-Google-TensorFlow_standard.jpg?alias=standard_600x400) for deep learning, would be great in order to organize and manage backends and frontends.

3 libs are candidates for me:
https://github.com/marmelab/ArchitectureTree (To see the whole sitemap)
https://github.com/marmelab/ng-admin (Etcd Json edition)
Demo for the Ng Admin (working only with JS): http://marmelab.com/ng-admin-demo/#/dashboard

For metrics, i saw an attempt with react.js but it is not working at all (https://github.com/bradgignac/vulcand/tree/bradgignac/dashboard/dashboard)

Would help a lot the get an outstanding proxy. Vulcan is all about metrics and micro-services.

PS. I do not know if you know Traefik (making too much general things for now)
https://github.com/emilevauge/traefik

Cheers,
Luc

[emmanuel]

Luc, thank you for the kind words, but I'm not seeing the value you're talking about 🤔. I haven't studied the implementation, but from my reading of the documentation, Romulus is basically what I had in mind when I started this project.

And now I want an Ingress controller, instead of annotations on a service :sigh:. Kubernetes (this whole space, really) is moving so fast, I am happy to deprecate this repo and adopt Romulus. I will be eager to evaluate a vulcand-based Ingress controller when one becomes available.

WDYT Luc?

On Dec 5, 2015, at 5:02 PM, lucmichalski [email protected] wrote:

Emmanuel,

At the opposite, I think that you provided an awesome piece of the puzzle in order to make of Vulcan a serious leading proxy for smart micro-services. So, I think that your repo is 30% of the whole as it own as it sticks to kubernetes as a boostrap template.

Maybe a more advanced tool would help Vulcan for sure. As I think that circuitbreakers, middlewares make me think about an web UI, link tensorflow (ex; http://images.csmonitor.com/csm/2015/11/946808_1_1109-Google-TensorFlow_standard.jpg?alias=standard_600x400) for deep learning, would be great in order to organize and manage backends and frontends.

3 libs are candidates for me:
https://github.com/marmelab/ArchitectureTree (To see the whole sitemap)
https://github.com/marmelab/ng-admin (Etcd Json edition)
Demo for the Ng Admin (working only with JS): http://marmelab.com/ng-admin-demo/#/dashboard

For metrics, i saw an attempt with react.js but it is not working at all (https://github.com/bradgignac/vulcand/tree/bradgignac/dashboard/dashboard)

Would help a lot the get an outstanding proxy. Vulcan is all about metrics and micro-services.

PS. I do not know if you know Traefik (making too much general things for now)
https://github.com/emilevauge/traefik

Cheers,
Luc

—
Reply to this email directly or view it on GitHub.

[ghost]

If you do not mind, maybe we should push your templates to Romulus, as it would be a great addition. (https://github.com/timelinelabs/romulus/issues/13)

My use case is really particular for vulcand as I am building a kind Image Recognition proxy with it. Some middlewares are doing pre-processing on the input pictures, and post-processing out the response.

So from my point of view, I prefer to route micro-services through annotation and be able to manage their replication on a micro level, only the most active (and make some dynamic backend loadbalancing like traefik does, not a simple round robin). So, I do not really need a classical load balancer with kubernetes deploy. benefits. I am pretty much chasing the scalability of deep learning frameworks with loads of micro-services/endpoints to query through a classical HTTP rest api.