Where and how do you deploy Apache OpenWhisk?

[Ben Browning <bb...@redhat.com>]

Over the past few months, a number of discussions and proposals on the
list have popped up on the list proposing varying degrees of
architectural changes to OpenWhisk. And, we have a large number of
deployment options with varying levels of maintenance.

With that in mind, I'd like to kick off a thread to gather some data
about OpenWhisk deployments in the wild. If you run some flavor of
OpenWhisk - whether straight Apache OpenWhisk or something based on it
- please answer as many of the questions below as you feel comfortable
doing. Please indicate whether your answers are for development,
testing, or production environments. Feel free to answer more than
once for different environments.

Where do you deploy OpenWhisk? On bare metal, VMs, Kubernetes, Mesos,
OpenShift, something else? In private datacenters? A cloud provider?

How do you deploy OpenWhisk? Related to the where, but what kind of
tooling do you use for the deployment itself?

Why did you choose those environments and deployment methods? Are you
satisfied with those choices?

Is there anything you'd like to see change in OpenWhisk to better
support your deployment and operational needs?

Is there anything else about your environments that would be helpful
for the community to know?


Thank you in advance!

Ben

Re: Where and how do you deploy Apache OpenWhisk?

[Dascalita Dragos <dd...@gmail.com>]

Hi Ben,
Thanks for starting this thread.

I'll chime in with my POV.

I see the deployment largely structured around 3 layers:

1. Container Management. This layer contains logic to spin up a Kubernetes
or Mesos cluster. It could be a managed cluster ( Azure Container Service,
Amazon ECS for Kubernetes, or others ) or a cluster installed using custom
scripts. The latter option gives more control, the first option is faster
to start with.

2. Install OpenWhisk components. This layers uses Helm or DCOS' Universe to
define packages, and kube or dcos CLI to install them.

3. Install system packages such as "whisk.system" actions. This layers uses
wskdeploy.

As you could probably guessed, each layer builds on the previous one.

To our particular setup:
- We use Ansible to orchestrate (1), (2), and (3).
- For (1) and (2) we have A/B deployments.
- (1) is installed with custom scripts to fully control the security in the
cluster. I.e. custom iptables rules isolating actions.



On Mon, Sep 17, 2018 at 7:17 AM Ben Browning <bb...@redhat.com> wrote:

> Over the past few months, a number of discussions and proposals on the
> list have popped up on the list proposing varying degrees of
> architectural changes to OpenWhisk. And, we have a large number of
> deployment options with varying levels of maintenance.
>
> With that in mind, I'd like to kick off a thread to gather some data
> about OpenWhisk deployments in the wild. If you run some flavor of
> OpenWhisk - whether straight Apache OpenWhisk or something based on it
> - please answer as many of the questions below as you feel comfortable
> doing. Please indicate whether your answers are for development,
> testing, or production environments. Feel free to answer more than
> once for different environments.
>
> Where do you deploy OpenWhisk? On bare metal, VMs, Kubernetes, Mesos,
> OpenShift, something else? In private datacenters? A cloud provider?
>
> How do you deploy OpenWhisk? Related to the where, but what kind of
> tooling do you use for the deployment itself?
>
> Why did you choose those environments and deployment methods? Are you
> satisfied with those choices?
>
> Is there anything you'd like to see change in OpenWhisk to better
> support your deployment and operational needs?
>
> Is there anything else about your environments that would be helpful
> for the community to know?
>
>
> Thank you in advance!
>
> Ben
>