Switch branches/tags
Nothing to show
Find file Copy path
325 lines (236 sloc) 13.2 KB

OpenWhisk Deployment on OpenShift

Build Status License

This repository contains the necessary templates and compatible docker images for deploying OpenWhisk on OpenShift.

What is Apache OpenWhisk

Apache OpenWhisk is a serverless, open source cloud platform that allows you to execute code in response to events at any scale. OpenWhisk handles the infrastructure and servers so you can focus on building amazing things.

Apache OpenWhisk allows developers to focus on writing value-adding code instead of burning hours on architecture and server management. Write in your preferred language to combine custom code with plug-and-play packages from our rich ecosystem of supporter services, and go live in hours instead of weeks.

Package creators can easily add their service to Apache OpenWhisk’s growing ecosystem to eliminate the need to build in-house solutions for third-party integrations, reach a broader community of developers, and increase adoption of their products and services.

The Apache OpenWhisk community is driven by open source contributors who are advancing this bleeding-edge technology, growing their skillsets, and pushing the boundaries of serverless technology.

For more information check the Apache OpenWhisk website.


The following command will deploy OpenWhisk in your OpenShift project using the latest ephemeral template in this repo:

oc process -f | oc create -f -

The shortened URL redirects to

This will take a few minutes. Eventually, all pods should enter the Running or Completed state, and the controller pod[s] should recognize the invoker pod[s] as ready. If you have wsk installed, run bin/ When it completes successfully, your cluster is ready.

Using wsk

Once your cluster is ready, you need to configure your wsk binary. If necessary, download it, unpack it, ensure it's in your PATH, and:

AUTH_SECRET=$(oc get secret whisk.auth -o yaml | grep "system:" | awk '{print $2}' | base64 --decode)
wsk property set --auth $AUTH_SECRET --apihost $(oc get route/openwhisk --template="{{}}")

That configures wsk to use your OpenWhisk. Use the -i option to avoid the validation error triggered by the self-signed cert in the nginx service.

wsk -i list
wsk -i action invoke /whisk.system/utils/echo -p message hello -b

If either fails, ensure you have the latest wsk installed.


The alarms package is not technically a part of the default OpenWhisk catalog, but since it's a simple way of experimenting with triggers and rules, we include a resource specification for it in our templates.

Try the following wsk commands:

wsk -i trigger create every-5-seconds \
    --feed  /whisk.system/alarms/alarm \
    --param cron '*/5 * * * * *' \
    --param maxTriggers 25 \
    --param trigger_payload "{\"name\":\"Odin\",\"place\":\"Asgard\"}"
wsk -i rule create \
    invoke-periodically \
    every-5-seconds \
wsk -i activation poll

Using minishift

First, start a recent version of minishift with ample resources:

minishift start --memory 8GB

And make the oc command available in your PATH:

eval $(minishift oc-env)

Assuming you have this repo cloned to your local workspace, run:


That will create an openwhisk project, install the resources from template.yml into it, and wait for all components to be ready. When it completes, you should have a functioning OpenWhisk platform, to which you can then point your wsk command.

You can do a quick smoke test of your cluster like so:


If you prefer not to clone this repo, you can simply follow the installation steps after creating a new project:

oc new-project openwhisk
oc process -f | oc create -f -

Testing Local Changes

If you'd like to test local changes you make to upstream OpenWhisk, e.g. the controller or invoker, first ensure you're using minishift's docker repo:

eval $(minishift docker-env)

Then when you build the OW images, override the prefix and tag:

./gradlew distDocker -PdockerImagePrefix=projectodd -PdockerImageTag=whatever

The projectodd prefix and whatever tag can be anything you like. You'll patch the running StatefulSets to refer to them so that any new pods they create will use your images.

# Patch the controller's StatefulSet
oc patch statefulset controller -p '{"spec":{"template":{"spec":{"containers":[{"name":"controller","image":"projectodd/controller:whatever"}]}}}}'

# Patch the invoker's StatefulSet
oc patch statefulset invoker -p '{"spec":{"template":{"spec":{"containers":[{"name":"invoker","image":"projectodd/invoker:whatever"}]}}}}'

# Now delete one or both pods to run your latest images
oc delete --force --now pod invoker-0 controller-0

With the StatefulSets patched, your build-test-debug cycle amounts to this: edit the source, run your distDocker task, e.g. core:controller:distDocker or core:invoker:distDocker with the above prefix/tag, and finally delete the relevant pod, e.g. controller-0 or invoker-0. This will trigger your patched StatefulSet to create a new pod with your changes.

Allow some time for the components to cleanly shutdown and rediscover themselves, of course. And while you're waiting, consider coming up with some good unit tests instead. ;)

And if you wish to publish your changes to DockerHub's projectodd organization:

COMMIT=$(git rev-parse HEAD | cut -c 1-7)
./gradlew distDocker -PdockerImagePrefix=projectodd -PdockerImageTag=$COMMIT


All of the OpenWhisk resources can be shutdown gracefully using the template. The -f parameter takes either a local file or a remote URL.

oc process -f template.yml | oc delete -f -
oc delete all -l template=openwhisk

Alternatively, you can just delete the project:

oc delete project openwhisk

Advanced Configuration

Persistent Data

If you'd like for data to survive reboots, there's a persistent-template.yml that will setup PersistentVolumeClaims.

Larger Clusters

There are some sensible defaults for larger persistent clusters in larger.env that you can use like so:

oc process -f persistent-template.yml --param-file=larger.env | oc create -f -

Performance Testing

Adjust the connection count and test duration of both below as needed. On a large system, be sure to test with connection counts in the hundreds.

With ab

For simple testing, use ab:

ab -c 5 -n 300 -k -m POST -H "Authorization: Basic $(oc get secret whisk.auth -o yaml | grep "system:" | awk '{print $2}')" "https://$(oc get route/openwhisk --template={{}})/api/v1/namespaces/whisk.system/actions/utils/echo?blocking=true&result=true"

With wrk

You can generate in-cluster load with wrk

echo -e "function main() {\n  return {body: 'Hello world'};\n}" > helloWeb.js
wsk -i action create helloWeb helloWeb.js --web=true
oc run -it --image williamyeh/wrk wrk --restart=Never --rm   --overrides='{"apiVersion":"v1", "spec":{"volumes":[{"name": "data", "emptyDir": {}}], "containers":[{"name": "wrk", "image": "williamyeh/wrk", "args": ["--threads", "4", "--connections", "50", "--duration", "30s", "--latency", "--timeout", "10s", "http://nginx/api/v1/web/whisk.system/default/helloWeb"], "volumeMounts": [{"mountPath": "/data", "name": "data"}]}]}}'

Activation statistics

The bin/ script can output throughput and waitTime numbers for recent function activations. This is useful when spot-checking overall system load and how long functions are waiting in queues inside OpenWhisk before being invoked.

Common Problems

Empty catalog (nothing from wsk list)

The following command should show a number of system packages:

wsk -i package list /whisk.system

If it doesn't, the install-catalog job probably failed. The first time you install OpenWhisk may take a very long time, due to the number of Docker images being pulled. This may cause the install-catalog job to give up, leaving you without the default system packages installed.

To remedy this, simply delete and recreate the job:

oc delete job install-catalog
oc process -f template.yml | oc create -f -

You'll see harmless AlreadyExists errors for all but the install-catalog job. Once its associated pod runs to completion, you should see output like the following:

$ wsk -i package list /whisk.system
/whisk.system/combinators                                              shared
/whisk.system/websocket                                                shared
/whisk.system/github                                                   shared
/whisk.system/utils                                                    shared
/whisk.system/slack                                                    shared
/whisk.system/samples                                                  shared
/whisk.system/watson-translator                                        shared
/whisk.system/watson-textToSpeech                                      shared
/whisk.system/watson-speechToText                                      shared
/whisk.system/weather                                                  shared
/whisk.system/alarms                                                   shared

The requested resource does not exist when creating an action

It might happen that when creating an action you get an error that the requested resource does not exist:

$ wsk -i action create md5hasher target/maven-java.jar --main org.apache.openwhisk.example.maven.App
error: Unable to create action 'md5hasher': The requested resource does not exist. (code 619)

If this happens, it could be that the API host is incorrect. So, start by inspecting the property values:

$ wsk property get
client cert
Client key
whisk auth                  789c46b1-...
whisk API host    
whisk API version           v1
whisk namespace             _
whisk CLI version           2018-02-28T21:13:48.864+0000
whisk API build             2018-01-01T00:00:00Z
whisk API build number      latest

API host should only contain the host name, no http:// in front. Fix it by resetting the API host:

$ wsk property set --apihost
ok: whisk API host set to

Now try adding the action again:

$ wsk -i action create md5hasher target/maven-java.jar --main org.apache.openwhisk.example.maven.App
ok: created action md5hasher


Apache OpenWhisk is an effort undergoing incubation at The Apache Software Foundation (ASF), sponsored by the Apache Incubator. Incubation is required of all newly accepted projects until a further review indicates that the infrastructure, communications, and decision making process have stabilized in a manner consistent with other successful ASF projects. While incubation status is not necessarily a reflection of the completeness or stability of the code, it does indicate that the project has yet to be fully endorsed by the ASF.