tm

README

A CLI for knative

Installation

There are two ways of tm CLI installation:

• Download a binary from the latest release

• Install CLI from sources:

git clone https://github.com/triggermesh/tm.git
cd tm
go install

Configuration

On TriggerMesh:

1. Request beta access to TriggerMesh cloud at https://triggermesh.com
2. Download your TriggerMesh configuration file by clicking on the download button in the upper right corner
3. Save the file as $HOME/.tm/config.json and you are ready to use the tm CLI

On your own knative cluster:

Assuming you have access to the Kubernetes API and have a working kubectl setup, tm should work out of the box.

Examples

Deploy service from Docker image

tm deploy service foo -f gcr.io/google-samples/hello-app:1.0 --wait

If you have Dockerfile for your service, you can use kaniko runtime to deploy it

tm deploy service foobar \
    -f https://github.com/knative/docs \
    --runtime https://raw.githubusercontent.com/triggermesh/knative-lambda-runtime/master/kaniko/runtime.yaml \
    --build-argument DIRECTORY=docs/serving/samples/hello-world/helloworld-go \
    --wait

or deploy service straight from Go source using OpenFaaS runtime

tm deploy service bar \
    -f https://github.com/golang/example \
    --runtime https://raw.githubusercontent.com/triggermesh/openfaas-runtime/master/go/openfaas-go-runtime.yaml \
    --build-argument DIRECTORY=hello \
    --wait

Moreover, for more complex deployments, tm CLI supports function definition parsing from YAML file and ability to combine multiple functions, runtimes and repositories

tm deploy -f https://github.com/tzununbekov/serverless

If you are interested in a building image without deploying knative service, then --build-only flag is available in "deploy service" command

Running Tests Locally

To run tests you first have to set namespace you have access to with the following command:

export NAMESPACE=yourNamespace

Run unit-tests with following command from project root directory:

make test

AWS Lambda

With the TriggerMesh CLI you can easily deploy AWS Lambda functions on Kubernetes:

Prepare local source for Golang function

mkdir lambda
cd lambda
cat > main.go <<EOF
package main

import (
        "fmt"
        "context"
        "github.com/aws/aws-lambda-go/lambda"
)

type MyEvent struct {
        Name string
}

func HandleRequest(ctx context.Context, name MyEvent) (string, error) {
        return fmt.Sprintf("Hello %s!", name.Name ), nil
}

func main() {
        lambda.Start(HandleRequest)
}
EOF

Deploy function using Go Knative Lambda Runtime

tm deploy service go-lambda -f . --runtime https://raw.githubusercontent.com/triggermesh/knative-lambda-runtime/master/go/runtime.yaml --wait

Lambda function available via http events

curl http://go-lambda.default.dev.triggermesh.io --data '{"Name": "Foo"}'
"Hello Foo!"

Here you can find more information about Knative Lambda Runtimes

Deployment pipelines

This feature is only available for Github.com repositories at the moment

With TriggerMesh CLI you can create fully functional deployment pipeline of existing git repository with a single command. In example below we're assuming that you have an access to Kubernetes cluster with Knative and Tekton pipelines installed. If you use TriggerMesh Cloud you should not worry about requirements; platform is ready to go.

As a first step, you should create new public repository in GitHub.com which we will use in our example. After the empty repository has been created, we need to push sample AWS Lambda project to it:

tm generate python foo
cd foo
git init
git add --all
git commit -m "Sample AWS Lambda project"
git remote add origin git@github.com:<USERNAME>/<REPOSITORY>.git
git push -u origin master

Now that we have repository with Python project, let's create build pipeline:

tm push | kubectl apply -f -

-this command creates several Knative and Tekton components:

1. Tekton task with tm image to build AWS Lambda project using KLR
2. Tekton taskrun to initiate project build and corresponding pipelineresource with source URL
3. TriggerMesh GitHub custom "third-party" containersource that allows to track events on GitHub repositories*
4. TriggerMesh Aktion transceiver and its configmap to create new taskruns on incoming events from GitHub containersource

* our GitHub containersource is aimed at simplifying event tracking and based on periodic GitHub API requests (one request per minute). As a result, you don't need to create and store any tokens. Downside of this approach is that containersource have requests rate limitation (60 requests per hour) and it doesn't work with private repositories. Both of these limitations can be bypassed by providing GitHub personal access token in push command parameter: tm push --token <TOKEN>

After few minutes you should be able to see new Knative service deployed in cluster. Any commits will trigger new build and deploy so that new function will reflect all code changes.

Docker registry

Docker images are used to run functions code in Knative services. This means that image registry is important part of service deployment scheme. Depending on type of service, Knative controller may either only pull or also push service image from and to registry. TriggerMesh CLI provides simple configuration interface to setup registry address and user access credentials.

Service from pre-build image

Most simple type of service deployment uses service based on pre-built Docker image available in public registry. This kind of service doesn't require any additional configuration and may be started with following command:

tm deploy service foo -f gcr.io/google-samples/hello-app:1.0 --wait

This case doesn't produce any images so no authentication is required.

If pre-built image stored in private registry, you must specify access credentials by running following command before starting deployment:

tm set registry-auth foo-registry

You will be asked to enter a registry address, username, and password - they will be saved to k8s secret and be used to pull images deployed under your service account.

Besides pulling, this secret may be used to push new images for service deployment based on function source code and build template. Name of one particular k8s secret should be passed to deployment command to make CLI work with private registry:

tm deploy service foo-private -f https://github.com/serverless/examples \
                              --runtime knative-node4-runtime \
                              --build-argument DIRECTORY=aws-node-serve-dynamic-html-via-http-endpoint \
                              --build-argument HANDLER=handler.landingPage \
                              --registry-secret foo-registry \
                              --wait

If user whose credentials are specified in foo-registry have "write" permissions, resulting service image will be pushed to URL composed as registry/username/service_name

Gitlab CI registry

TriggerMesh CLI can be used as deployment step in GitLab CI pipeline, but considering tokens security policy, user must manually create CI deployment token as described here. Deployment token must have registry read permission and should be valid for as long as the service expected to be active. If token is created, tm deployment step must include following commands:

...
script:
  - tm -n "$KUBE_NAMESPACE" set registry-auth gitlab-registry --registry "$CI_REGISTRY" --username "$CI_REGISTRY_USER" --password "$CI_JOB_TOKEN" --push
  - tm -n "$KUBE_NAMESPACE" set registry-auth gitlab-registry --registry "$CI_REGISTRY" --username "$CI_DEPLOY_USER" --password "$CI_DEPLOY_PASSWORD" --pull
...

After this, you may pass --registry-secret gitlab-registry parameter to tm deploy command (or in serverless.yml) so that Knative could authenticate against Gitlab registry. GitLab registry doesn't provide permanent read-write token that can be used in CI, but it has job-specific CI_JOB_TOKEN with "write" permission which is valid only while CI job running and CI_DEPLOY_PASSWORD with read permission which we created before. Considering this, we can see that CLI set registry-auth command supports --push and --pull flags that indicates which secret must be used to push image and which for "pull" operations only. Resulting images will be stored under registry.gitlab.com/username/project/function_name path

Custom registry name

While using a username as a registry identifier (docker.io/username) is a common practice, in some cases we must be able to use different values for an authentication and in destination URL (for example, gcr.io). TriggerMesh CLI set registry-auth command provides such ability by exposing an optional --project argument which will be used as a part of the image URL instead of the username:

TOKEN=$(gcloud auth print-access-token)
tm set registry-auth gcr --registry eu.gcr.io --project my-org/my-project --username oauth2accesstoken --password $TOKEN
tm generate python
tm deploy -f python --registry-secret gcr --wait

As a result, Knative service image will be pushed to eu.gcr.io/my-org/my-project registry

AWS ECR

ECR is a specific case of the custom registry destination with an additional requirement - a repository must be created before pushing the image. Thus, service deployment steps should be slightly altered:

1. Create ECR repository in <project>/<service> format where "project" is an arbitrary identifier for the service (e.g., namespace) and "service" is the name of the service that is being deployed (in the example below it is python-test).

2. Retrieve ECR token:

   TOKEN=$(aws ecr get-login-password --region <region>)
   

3. Create the registry auth secret by running following command:

   tm set registry-auth ecr --registry <registry host> --project <project> --username AWS --password $TOKEN
   

4. Deploy the service. For example Python KLR:

   tm deploy service python-test -f https://github.com/serverless/examples \
     --runtime https://raw.githubusercontent.com/triggermesh/knative-lambda-runtime/master/python37/runtime.yaml \
     --registry-secret ecr \
     --build-argument DIRECTORY=aws-python-simple-http-endpoint \
     --build-argument HANDLER=handler.endpoint \
     --wait
   

Unauthenticated registry

Besides hosted registries, the TriggerMesh CLI may work with unauthenticated registries which do not require setting access credentials. For such cases, you may simply add --registry-host argument to the deployment command with registry domain name parameter and the resulting image will be pushed to registry-host/namespace/service_name URL

How-To

For additional details on how to use tm, consult the HOWTO Guide

Support

We would love your feedback on this CLI tool so don't hesitate to let us know what is wrong and how we could improve it, just file an issue

Code of Conduct

This plugin is by no means part of CNCF but we abide by its code of conduct