EventListener is a Kubernetes custom resource that allows users a declarative way to process incoming HTTP based events with JSON payloads. EventListeners expose an addressable "Sink" to which incoming events are directed. Users can declare TriggerBindings to extract fields from events, and apply them to TriggerTemplates in order to create Tekton resources. In addition, EventListeners allow lightweight event processing using Event Interceptors.
- EventListener
To define a configuration file for an EventListener resource, you can specify
the following fields:
- Required:
apiVersion- Specifies the API version, for exampletriggers.tekton.dev/v1alpha1.kind- Specifies theEventListenerresource object.metadata- Specifies data to uniquely identify theEventListenerresource object, for example aname.spec- Specifies the configuration information for your EventListener resource object. In order for an EventListener to do anything, the spec must include:triggers- Specifies a list of Triggers to runserviceAccountName- Specifies the ServiceAccount that the EventListener uses to create resources
- Optional:
serviceType- Specifies what type of service the sink pod is exposed asreplicas- Specifies the number of EventListener podspodTemplate- Specifies the PodTemplate for your EventListener podresources- Specifies the Kubernetes Resource information for your EventListener podnamespaceSelector- Specifies the namespaces where EventListener can fetch triggers from and create Tekton resources.
The serviceAccountName field is required. The ServiceAccount that the
EventListener sink uses to create the Tekton resources.
The ServiceAccount needs a Role that with "get", "list", and "watch" verbs for each Triggers resource as well as a ClusterRole with read access to ClusterTriggerBindings. In addition, it needs to have "create"
permissions on the Pipeline resources it needs to create. See a working example at ../examples/rbac.yaml.
If your EventListener is using namespaceSelectors, the ServiceAccount will require a Cluster role to have read permissions for all Triggers resources across the cluster.
The triggers field is required. Each EventListener can consist of one or more
triggers. A Trigger consists of:
name- (Optional) a valid Kubernetes nameinterceptors- (Optional) list of interceptors to usebindings- (Optional) A list of bindings to use. Can either be a reference to existingTriggerBindingresources or embedded name/value pairs.template- (Optional) Either a reference to a TriggerTemplate object or an embedded TriggerTemplate spec.triggerRef- (Optional) Reference to theTrigger.
A trigger field must either have a template (along with needed bindings and interceptors) or a reference to another Trigger using triggerRef.
triggers:
- name: trigger-1
interceptors:
- github:
eventTypes: ["pull_request"]
bindings:
- ref: pipeline-binding # Reference to a TriggerBinding object
- name: message # Embedded Binding
value: Hello from the Triggers EventListener!
template:
ref: pipeline-templateOr with only triggerRef:
triggers:
- triggerRef: triggerOr with an embedded triggerTemplate spec:
triggers:
- name: "my-trigger"
template:
spec:
params:
- name: "my-param-name"
resourceTemplates:
- apiVersion: "tekton.dev/v1beta1"
kind: TaskRun
metadata:
generateName: "pr-run-"
spec:
taskSpec:
steps:
- image: ubuntu
script: echo "hello there"Also, to support multi-tenant styled scenarios, where an administrator may not want all triggers to have
the same permissions as the EventListener, a service account can optionally be set at the trigger level
and used if present in place of the EventListener service account when creating resources:
triggers:
- name: trigger-1
serviceAccount:
name: trigger-1-sa
namespace: event-listener-namespace
interceptors:
- github:
eventTypes: ["pull_request"]
bindings:
- ref: pipeline-binding
- ref: message-binding
template:
ref: pipeline-templateAn update to the Role assigned to the EventListener's SeviceAccount is needed to allow it to impersonate
the ServiceAccount specified for the trigger.
rules:
- apiGroups: [""]
resources: ["serviceaccounts"]
verbs: ["impersonate"]The serviceType field is deprecated. Consider using Resources field.
The serviceType field is optional. EventListener sinks are exposed via
Kubernetes Services.
By default, the serviceType is ClusterIP which means any pods running in the
same Kubernetes cluster can access services' via their cluster DNS. Other valid
values are NodePort and LoadBalancer. Check the
Kubernetes Service types
documentations for details.
For external services to connect to your cluster (e.g. GitHub sending webhooks), check out the guide on exposing EventListeners.
The replicas field is optional. By default, the number of replicas of EventListener is 1.
If you want to deploy more than one pod, you can specify the number to replicas field.
Note: If user sets replicas field while creating/updating eventlistener yaml then it won't respects replicas values edited by user manually on deployment or through any other mechanism (ex: HPA).
The podTemplate field is deprecated. Consider using Resources field.
The podTemplate field is optional. A PodTemplate is specifications for
creating EventListener pod. A PodTemplate consists of:
tolerations- list of toleration which allows pods to schedule onto the nodes with matching taints. This is needed only if you want to schedule EventListener pod to a tainted node.nodeSelector- key-value labels the node has which an EventListener pod should be scheduled on.
spec:
podTemplate:
nodeSelector:
app: test
tolerations:
- key: key
value: value
operator: Equal
effect: NoScheduleThe resources field is optional.
Resource field helps to provide Kubernetes or custom resource information.
For more info on the design refer TEP-0008
Right now the resources field is optional in order to support backward compatibility with original behavior of podTemplate, serviceType and serviceAccountName fieds.
In the future, we plan to remove serviceType and podTemplate from the EventListener spec in favor of the resources field.
For now resources has support for kubernetesResource but later it will have a support for Custom CRD(ex: Knative Service) as customResource
kubernetesResource have two fields
- ServiceType
- Spec(PodTemplateSpec)
spec:
resources:
kubernetesResource:
serviceType: NodePort
spec:
template:
metadata:
labels:
key: "value"
annotations:
key: "value"
spec:
serviceAccountName: tekton-triggers-github-sa
nodeSelector:
app: test
tolerations:
- key: key
value: value
operator: Equal
effect: NoScheduleWith the help of kubernetesResource user can specify PodTemplateSpec.
Right now the allowed values as part of podSpec are
ServiceAccountName
NodeSelector
Tolerations
Volumes
Containers
- Resources
- VolumeMounts
- Env
EventListener sinks are exposed as Kubernetes services that are backed by a Pod
running the sink logic. The logging configuration can be controlled via the
config-logging-triggers ConfigMap present in the namespace that the
EventListener was created in. This ConfigMap is automatically created and
contains the default values defined in
config-logging.yaml.
To access logs for the EventListener sink, you can query for pods with the
eventlistener label set to the name of your EventListener resource:
kubectl get pods --selector eventlistener=my-eventlistenerThe namespaceSelector field is optional.
This field determines the namespaces where EventListener can search for triggers and
create Tekton resources. If this field isn't provided, EventListener will only serve Triggers from its
own namespace.
Snippet below will function in foo and bar namespaces.
namespaceSelector:
matchNames:
- foo
- barIf EventListener is required to listen to serve the whole cluster, then below snippet
can be used where we only provide single argument for matchNames as *.
namespaceSelector:
matchNames:
- *By default, EventListeners will attach the following labels automatically to all resources it creates:
| Name | Description |
|---|---|
| triggers.tekton.dev/eventlistener | Name of the EventListener that generated the resource. |
| triggers.tekton.dev/trigger | Name of the Trigger that generated the resource. |
| triggers.tekton.dev/eventid | UID of the incoming event. |
Since the EventListener name and Trigger name are used as label values, they must adhere to the Kubernetes syntax and character set requirements for label values.
All the annotations provided in Eventlistener will be further propagated to the service and deployment created by that Eventlistener.
Example: You may need to add some annotation to the service like you need to annotation for TLS support to a LoadBalancer service on AWS, you can specify that annotation in Eventlistener and it will be available to the service created by EventListener.
apiVersion: triggers.tekton.dev/v1alpha1
kind: EventListener
metadata:
name: eventlistener
annotations:
service.beta.kubernetes.io/aws-load-balancer-backend-protocol: https
Note: If there are any annotations attached to the service or the deployment, they will get overwritten by the annotations available in the eventlistener.
Triggers within an EventListener can optionally specify interceptors, to
modify the behavior or payload of Triggers.
Event Interceptors can take several different forms today:
- Webhook Interceptors
- GitHub Interceptors
- GitLab Interceptors
- Bitbucket Interceptors
- CEL Interceptors
Webhook Interceptors allow users to configure an external k8s object which
contains business logic. These are currently specified under the Webhook
field, which contains an
ObjectReference
to a Kubernetes Service. If a Webhook Interceptor is specified, the
EventListener sink will forward incoming events to the service referenced by
the Interceptor over HTTP. The service is expected to process the event and
return a response back. The status code of the response determines if the
processing is successful - a 200 response means the Interceptor was successful
and that processing should continue, any other status code will halt Trigger
processing. The returned request (body and headers) is used as the new event
payload by the EventListener and passed on the TriggerBinding. An Interceptor
has an optional header field with key-value pairs that will be merged with event
headers before being sent;
canonical names
must be specified.
The incoming request URL (received by the EventListener) is provided in the
Eventlistener-Request-URL header provided to the Webhook interceptor.
When multiple Interceptors are specified, requests are piped through each Interceptor sequentially for processing - e.g. the headers/body of the first Interceptor's response will be sent as the request to the second Interceptor. It is the responsibility of Interceptors to preserve header/body data if desired. The response body and headers of the last Interceptor is used for resource binding/templating.
To be an Event Interceptor, a Kubernetes object should:
- Be fronted by a regular Kubernetes v1 Service over port 80
- Accept JSON payloads over HTTP
- Accept HTTP POST requests with JSON payloads.
- Return a HTTP 200 OK Status if the EventListener should continue processing the event
- Return a JSON body back. This will be used by the EventListener as the event payload for any further processing. If the Interceptor does not need to modify the body, it can simply return the body that it received.
- Return any Headers that might be required by other chained Interceptors or any bindings.
Note: It is the responsibility of Interceptors to preserve header/body data if desired. The response body and headers of the last Interceptor is used for resource binding/templating.
---
apiVersion: triggers.tekton.dev/v1alpha1
kind: EventListener
metadata:
name: listener-interceptor
spec:
serviceAccountName: tekton-triggers-example-sa
triggers:
- name: foo-trig
interceptors:
- webhook:
header:
- name: Foo-Trig-Header1
value: string-value
- name: Foo-Trig-Header2
value:
- array-val1
- array-val2
objectRef:
kind: Service
name: gh-validate
apiVersion: v1
namespace: default
bindings:
- ref: pipeline-binding
template:
ref: pipeline-templateGitHub Interceptors contain logic to validate and filter webhooks that come from GitHub. Supported features include validating webhooks actually came from GitHub using the logic outlined in GitHub documentation, as well as filtering incoming events.
To use this Interceptor as a validator, create a secret string using the method
of your choice, and configure the GitHub webhook to use that secret value.
Create a Kubernetes secret containing this value, and pass that as a reference
to the github Interceptor.
To use this Interceptor as a filter, add the event types you would like to
accept to the eventTypes field. Valid values can be found in GitHub
docs.
The body/header of the incoming request will be preserved in this Interceptor's response.
triggers:
- name: github-listener
interceptors:
- github:
secretRef:
secretName: github-secret
secretKey: secretToken
eventTypes:
Check out a full example of using GitHub Interceptor in examples/github
GitLab Interceptors contain logic to validate and filter requests that come from GitLab. Supported features include validating that a webhook actually came from GitLab, using the logic outlined in GitLab documentation, and to filter incoming events based on the event types. Event types can be found in GitLab documentation.
To use this Interceptor as a validator, create a secret string using the method
of your choice, and configure the GitLab webhook to use that secret value.
Create a Kubernetes secret containing this value, and pass that as a reference
to the gitlab Interceptor.
To use this Interceptor as a filter, add the event types you would like to
accept to the eventTypes field.
The body/header of the incoming request will be preserved in this Interceptor's response.
apiVersion: triggers.tekton.dev/v1alpha1
kind: EventListener
metadata:
name: gitlab-listener-interceptor
spec:
serviceAccountName: tekton-triggers-example-sa
triggers:
- name: foo-trig
interceptors:
- gitlab:
secretRef:
secretName: foo
secretKey: bar
eventTypes:
- Push Hook
bindings:
- ref: pipeline-binding
template:
ref: pipeline-templateThe Bitbucket interceptor provides support for hooks originating in Bitbucket server, providing server hook signature validation and event-filtering. Bitbucket cloud is not currently supported by this interceptor, as it has no secret validation, so you could match on the incoming requests using the CEL interceptor.
To use this Interceptor as a validator, create a secret string using the method
of your choice, and configure the Bitbucket webhook to use that secret value.
Create a Kubernetes secret containing this value, and pass that as a reference
to the bitbucket Interceptor.
To use this Interceptor as a filter, add the event types you would like to
accept to the eventTypes field. Valid values can be found in Bitbucket
docs.
The body/header of the incoming request will be preserved in this Interceptor's response.
---
apiVersion: triggers.tekton.dev/v1alpha1
kind: EventListener
metadata:
name: bitbucket-listener
spec:
serviceAccountName: tekton-triggers-bitbucket-sa
triggers:
- name: bitbucket-triggers
interceptors:
- bitbucket:
secretRef:
secretName: bitbucket-secret
secretKey: secretToken
eventTypes:
- repo:refs_changed
bindings:
- ref: bitbucket-binding
template:
ref: bitbucket-templateCEL Interceptors can be used to filter or add extra information to incoming events, using the CEL expression language.
Please read the cel-spec language definition for more details on the expression language syntax.
The cel-trig-with-matches trigger below filters events that don't have an
'X-GitHub-Event' header matching 'pull_request'.
It also modifies the incoming event, adding an extra key to the JSON body, with a truncated string coming from the hook body.
triggers:
- name: cel-trig-with-matches
interceptors:
- cel:
filter: "header.match('X-GitHub-Event', 'pull_request')"
overlays:
- key: truncated_sha
expression: "body.pull_request.head.sha.truncate(7)"
bindings:
- name: sha
value: $(extensions.truncated_sha)In addition to the standard expressions provided by CEL, Triggers supports some useful functions for dealing with event data CEL expressions.
The filter expression must return a true value if this trigger is to be
processed, and the overlays applied.
Optionally, no filter expression can be provided, and the overlays will be
applied to the incoming body.
apiVersion: triggers.tekton.dev/v1alpha1
kind: EventListener
metadata:
name: cel-eventlistener-no-filter
spec:
serviceAccountName: tekton-triggers-example-sa
triggers:
- name: cel-trig
interceptors:
- cel:
overlays:
- key: extensions.truncated_sha
expression: "body.pull_request.head.sha.truncate(7)"
bindings:
- ref: pipeline-binding
template:
ref: pipeline-templateThe CEL interceptor supports "overlays", these are CEL expressions whose values are added to the incoming event under a
top-level extensions field and are accessible by TriggerBindings.
triggers:
- name: cel-trig
interceptors:
- cel:
overlays:
- key: extensions.truncated_sha
expression: "body.pull_request.head.sha.truncate(7)"
- key: extensions.branch_name
expression: "body.ref.split('/')[2]"In this example, the bindings will see two added" fields -- extensions.truncated_sha and extensions.branch_name in
addition to the regular body and header fields.
The key element of the overlay can create new elements or replace existing elements within the extensions field.
Note that the original incoming body is not modified. Instead, new fields are added to a separate top-level extensions
field that is accessible by TriggerBindings.
For example, this expression:
- key: short_sha
expression: "truncate(body.pull_request.head.sha, 7)"Would see the short_sha being created in the extensions field:
{
"body": {
"ref": "refs/heads/master",
"pull_request": {
"head": {
"sha": "6113728f27ae82c7b1a177c8d03f9e96e0adf246"
}
}
},
"extensions": {
"short_sha": "6113728"
}
}It's even possible to replace existing fields, by providing a key that matches the path to an existing value.
Anything that is applied as an overlay can be extracted using a binding e.g.
apiVersion: triggers.tekton.dev/v1alpha1
kind: TriggerBinding
metadata:
name: pipeline-binding-with-cel-extensions
spec:
params:
- name: gitrevision
value: $(extensions.branch_name)
- name: branch
value: $(extensions.short_sha)This section documents the current behavior for passing data between interceptors. This will change as we fully implement #271.
CEL Interceptor: Overlays from the CEL interceptor do not modify the event body. Instead they add fields to the top level extensions field.
Webhook Interceptors: Webhook Interceptors can modify the event body currently. However, this will change in a future release when they will have to write to the extensions field like the CEL interceptor. Since the webhook interceptor does not have access to the top level extensions field, the EventListener will add the extensions field to the body before sending it to the webhook interceptor. As an example, consider the following interceptor chain:
interceptors:
- cel:
overlays:
- key: "truncated_sha"
expression: "body.sha.truncate(5)"
- webhook:
objectRef:
kind: Service
name: some-interceptor
apiVersion: v1
- cel:
filter: "body.extensions.truncated_sha == \"abcde\"" # Can also be extensions.truncated_sha == \"abcde\"In the above snipped, the first CEL interceptor adds the truncated_sha field. To ensure the following webhook interceptor can use this field, the EventListener will add it to the body. So, the body received by the webhook interceptor will look as follows:
{
"sha": "abcdefghi", // Original field
"extensions": {
"truncated_sha": "abcde"
}
}
Assuming the webhook interceptor does not then modify the body, the last CEL interceptor (as well as any bindings) will have access to truncated_sha both via the body as well as via extensions i.e both $(body.extensions.truncated_sha) as well as $(extensions.truncated_sha)
The EventListener responds with 201 Created status code when at least one of the trigger is executed successfully. Otherwise, it returns 202 Accepted status code. The EventListener responds with following message after receiving the event:
{"eventListener":"listener","namespace":"default","eventID":"h2bb7"}eventListener- Refers to the EventListener Name.namespace- Refers to the namespace of the EventListenereventID- Refers to the uniqueID that gets assigned to each incoming request
Lets understand how an EventListener works with an example using GitHub
- Create a sample GitHub example
kubectl create -f https://github.com/tektoncd/triggers/tree/master/examples/github- Once the EventListener is created, the Triggers controller will create a new
DeploymentandServicefor the EventListener. We can usekubectlto see them running:
kubectl get deployment
NAME READY UP-TO-DATE AVAILABLE AGE
el-github-listener-interceptor 1/1 1 1 11s
kubectl get svc
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
el-github-listener-interceptor ClusterIP 10.99.188.140 <none> 8080/TCP 52sThe Triggers controller uses fields from the EventListener's spec (which is described in the Syntax section, as well as metadata.labels
in addition to some pre-configured information like (container Image, Name, Port) to create the Deployment and Service.
We follow a naming convention while creating these resources. An EventListener named foo will create a deployment and a service both named el-foo.
Once all the resources are up and running user can get a URL to send webhook events. This URL points to the service created above and points to the deployment.
kubectl get eventlistener
NAME ADDRESS AVAILABLE REASON
github-listener-interceptor http://el-github-listener-interceptor.ptest.svc.cluster.local:8080 True MinimumReplicasAvailableFollow GitHub example to try out locally.
For complete examples, see the examples folder.
Eventlistener sink timeout if EventListener isn't able to process the request within certain duration. The response timeout configuration is defined in controller.yaml.
-el-readtimeout: This define ReadTimeOut for sink server. Default value is 5 s.-el-writetimeout: This define WriteTimeout for sink server. Default value is 40s.-el-idletimeout: This define the IdleTimeout for sink server. Default value is 120s.-el-timeouthandler: This define the Timeout for Handler for sink server's route. Default value is 30s.
The EventListener is effectively an additional form of client into Tekton, versus what
example usage via kubectl or tkn which you have seen elsewhere. In particular, the HTTP based
events bypass the normal Kubernetes authentication path you get via kubeconfig files
and the kubectl config family of commands.
As such, there are set of items to consider when deciding how to
- best expose (each) EventListener in your cluster to the outside world.
- best control how (each) EventListener and the underlying API Objects described below access, create, and update Tekton related API Objects in your cluster.
Minimally, each EventListener has its ServiceAccountName as noted below and all events coming over the "Sink" result in any Tekton resource interactions being done with the permissions assigned to that ServiceAccount.
However, if you need differing levels of permissions over a set of Tekton resources across the various Triggers and Interceptors, where not all Triggers or Interceptors can manipulate certain Tekton Resources in the same way, a simple, single EventListener will not suffice.
Your options at that point are as follows:
You can create multiple EventListener objects, where your set of Triggers and Interceptors are spread out across the EventListeners.
If you create each of those EventListeners in their own namespace, it becomes easy to assign varying permissions to the ServiceAccount of each one to serve your needs. And often times namespace creation is coupled with a default set of ServiceAccounts and Secrets that are also defined. So conceivably some administration steps are taken care of. You just update the permissions of the automatically created ServiceAccounts.
Possible drawbacks:
- Namespaces with associated Secrets and ServiceAccounts in an aggregate sense prove to be the most expensive
items in Kubernetes underlying
etcdstore. In larger clustersetcdstorage capacity can become a concern. - Multiple EventListeners means multiple HTTP ports that must be exposed to the external entities accessing
the "Sink". If you happen to have a HTTP Firewall between your Cluster and external entities, that means more
administrative cost, opening ports in the firewall for each Service, unless you can employ Kubernetes
Ingressto serve as a routing abstraction layer for your set of EventListeners.
Multiple EventListeners per namespace will most likely mean more ServiceAccount/Secret/RBAC manipulation for the administrator, as some of the built in generation of those artifacts as part of namespace creation are not applicable.
However you will save some on the etcd storage costs by reducing the number of namespaces.
Multiple EventListeners and potential Firewall concerns still apply (again unless you employ Ingress).
Being able to set a ServiceAccount on an EventListenerTrigger allows for finer grained permissions as well.
You still have to create the additional ServiceAccounts.
But staying within 1 namespace and minimizing the number of EventListeners with their associated "Sinks" minimizes
concerns around etcd storage and port considerations with Firewalls if Ingress is not utilized.
Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License.
Triggers now support both HTTP and HTTPS connection by adding few configuration to eventlistener.
To setup TLS connection add two set of reserved environment variables TLS_CERT and TLS_KEY using secretKeyRef env type
where we need to specify the secret which contains cert and key files. See the full example for more details.
Refer TEP-0027 for more information on design and user stories.