openwhisk-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Carlos Santana <csantan...@gmail.com>
Subject Re: Proposal for OpenWhisk + Kubernetes integration
Date Fri, 17 Mar 2017 00:22:25 GMT
Hi Daniel Lavine

   My recommendation for now, would be to avoid working on branch (copy of
whisk) and instead create a folder in the dev-tools repo "kubernetes" have
a script there that git clones main repo of openwhisk , and then tried to
use it as is build the docker images, and then deploy to kubernetes, maybe
you need ansible or maybe not, I'm not an expert on kubernetes.

If you have modifications delta to main scala code for example for
controller or invoker, then have a patch or delta that you put on top of
what you clone from the main repo.

We are trying to remove concerns from the main repo not add more, to leave
the main repo with core container code, and removing things like tooling
and deployment, for example we remove the catalog which is a deployment
concern not main core containers, also plan to remove cli tooling.

If you have working scripts/code you can start submitting PRs to the
dev-tools and would be easier to experiment, and for folks to give it a try
like what we already have for docker-compose, there is also interest on
other type of deployments like DC/OS can also go there, or chef or puppet,
the points that we can experiment with different deployments options in the
repo making the core containers common across all those deployment
scenarios.

You can also can create an issue in the dev-tools repo to list the ideas,
tasks, issues and this can become a zenHub Epic with children as you and
other get stuff working

Once we get something up and running we start looking best way to make
changes to main components to adjust to different deployment targets, for
example today we use in ansible "-e mode=deploy" like in "ansible-playbook
-e mode=deploy" we could use that environment variable mode
(mode=deploy_k8s), or another one to select what configuration settings
would be different from the normal deployment we have today, or pass a
variable or settings via consul or environment variables to controller, or
invoker to run  certain code path for the deployment type.

It will cool to see that repo be a collection of all the different ways
people can install openwhisk, we could even move vagrant there, or
instructions on how to deploy to different VM providers (digital ocean,
linode, softlayer, etc..) for development/debug or for full operations. Now
that I think about it maybe the openstack distribute deployment we have in
the main repo could move here also and have more visibility.

[1] https://github.com/openwhisk/openwhisk-devtools


--Carlos




On Thu, Mar 16, 2017 at 7:50 PM Daniel Lavine <dlavine@us.ibm.com> wrote:

> Hey Dan,
>
> We are working on a fork of the OpenWhisk project and are extending the
> Ansible orchestration so that it can issue Kubernetes commands. We have
> created a Dockerfile which pulls in the OpenWhisk code and installs all of
> the required dependencies to run the Ansible play books. With this image we
> should then be able to create a Kube job that sets up OpenWhisk from within
> a Kubernets environment. This gives us the ability to leverage things like
> Kube-DNS when Ansible goes and performs any kind of restful configuration
> (i.e. seeding consul) and a fairly consistent way of working with a
> Kubernetes cluster.
>
> Currently, we are using the built Docker image as a pod and deploying that
> pod into Kubernets. From there we exec into that pod and are running the
> Ansible playbooks by hand. Then extending any of the command to use Kube
> commands whenever we need to.
>
> Ideally if this approach goes well and the solution make sense, then we
> would like to contribute this back into the main OpenWhisk branch.
>
> Lastly, we are doing a bit more work on ironing out our proposal and the
> goals we would like to hit so that they are a clear and measurable piece of
> work. Once that is clear, then we will post it to the wiki you created.
>
> Dan Lavine
>
>
> [image: Inactive hide details for Daniel Krook---03/16/2017 04:08:14
> AM---Hi Dan, As you can see from that issue Rodric referenced, I'v]Daniel
> Krook---03/16/2017 04:08:14 AM---Hi Dan, As you can see from that issue
> Rodric referenced, I've also being looking
>
> From: Daniel Krook/White Plains/IBM@IBMUS
> To: dev@openwhisk.apache.org
> Date: 03/16/2017 04:08 AM
> Subject: Re: Proposal for OpenWhisk + Kubernetes integration
> ------------------------------
>
>
>
>
> Hi Dan,
>
> As you can see from that issue Rodric referenced, I've also being looking
> into K8s and often hear others express interest as well [1]. It will be
> great to have people working together on this.
>
> Here is a wiki page I started where we can capture the proposal:
> https://cwiki.apache.org/confluence/display/OPENWHISK/OpenWhisk+on
> +Kubernetes
>
> I also created a Slack channel if you want to chat and invite others, we'll
> definitely want to loop in the folks from Hyper.sh and StackPoint.
> https://openwhisk-team.slack.com/messages/kubernetes/
>
> I'd like to hear more about what you've built already.
>
>
> [1] https://github.com/openwhisk/openwhisk/issues/1402
>
>
>
> Daniel Krook
> http://krook.info/
>
>
>
> > From: Daniel Lavine
> >
> > Hi everyone,
> >
> > My name is Dan Lavine and I'm a long time contributor to the Cloud
> Foundry
> > Open source project and have an interest in other distributed
> cloud-native
> > platforms like OpenWhisk and Kubernetes.
> >
> > Given the growing interest in serverless architectures and container
> > orchestration technologies, I think integrating OpenWhisk with Kube could
> > be a cool project and worthwhile effort. My colleague, Kalonji Bankole
> and
> > I have have been working together to get some stuff up and running, but
> > there is a ways to go.
> >
> > If others are interested, we would like to share what we currently have.
> > Documenting what our current approach is and get feedback for what we are
> > working on would be a huge help.
> >
> > Does anyone have interest in this work?
> >
> > I look forward to hearing back from you!
> >
> > Dan Lavine
>
>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message