Image resources provide a configuration for kpack to build and maintain a docker image utilizing Cloud Native Buildpacks. kpack will monitor the inputs to the image resource to rebuild the OCI image when the underlying source or the builder's buildpacks or stacks have changed.
Corresponding kp
cli command docs here.
The following defines the relevant fields of the image
resource spec in more detail:
tag
: The image tag.additionalTags
: Any additional list of image tags that should be published. This list of tags is mutable.builder
: Configuration of thebuilder
resource the image builds will use. See more info Builder Configuration.serviceAccountName
: The Service Account name that will be used for credential lookup.source
: The source code that will be monitored/built into images. See the Source Configuration section below.cache
: Caching configuration, two variants are available:volume.size
: Creates a Volume Claim of the given sizevolume.storageClassName
: (Optional) Creates a Volume Claim of the given storageClassName. If unset, the default storage class is used. The field is immutable.registry.tag
: Creates an image with cached contents
failedBuildHistoryLimit
: The maximum number of failed builds for an image that will be retained.successBuildHistoryLimit
: The maximum number of successful builds for an image that will be retained.imageTaggingStrategy
: Allow for builds to be additionally tagged with the build number. Valid options areNone
andBuildNumber
.build
: Configuration that is passed to every image build. See Build Configuration section below.defaultProcess
: The default process type for the built OCI imageprojectDescriptorPath
: Path to the project descriptor file relative to source root dir orsubPath
if set. If unset, kpack will look forproject.toml
at the root dir orsubPath
if set.cosign
: Configuration for additional cosign image signing. See Cosign Configuration section below.
The tag
field is the location the built OCI image will be written to for each build. This field is immutable.
Examples:
tag: dockerhubuser/repo
tag: dockerhubuser/repo:my-image
tag: gcr.io/project/repo
The additionalTags
is a list of locations the built OCI image will be written to in addition to the tag
. Additional tags must be in the same registry as the tag
. Cross registry exporting is not supported. This field can be modified.
Example:
tag: my-registry.io/project/repo
additionalTags:
- my-registry.io/project/repo:some-version
- my-registry.io/project/repo:some-metadata
- my-registry.io/project/other-repo
The builder
field describes the builder resource that will build the OCI images for a provided image configuration. It can be defined in exactly one of the following ways:
-
Cluster Builder
builder: name: cluster-builder-name kind: ClusterBuilder
name
: The name of the ClusterBuilder resource in kubernetes.kind
: The type as defined in kubernetes. This will always be ClusterBuilder.
-
Namespaced Builder
builder: name: builder-name kind: Builder
name
: The name of the Builder resource in kubernetes.kind
: The type as defined in kubernetes. This will always be Builder.
Note: This image resource can only reference builders defined in the same namespace. This is not true for ClusterBuilders because they are not namespace scoped.
The source
field is a composition of a source code location and a subpath
. It can be configured in exactly one of the following ways:
-
Git
source: git: url: "" revision: "" initializeSubmodules: false subPath: ""
git
: (Source Code is a git repository)url
: The git repository url. Both https and ssh formats are supported; with ssh format requiring a ssh secret.revision
: The git revision to use. This value may be a commit sha, branch name, or tag.initializeSubmodules
: Initialize submodules inside repo, recurses up to a max depth of 10 submodules.
subPath
: A subdirectory within the source folder where application code resides. Can be ignored if the source code resides at theroot
level.
-
Blob
source: blob: url: "" stripComponents: 0 subPath: ""
blob
: (Source Code is a blob/jar in a blobstore)url
: The URL of the source code blob. This blob needs to either be publicly accessible or have the access token in the URLstripComponents
: Optional number of directory components to strip from the blobs content when extracting.
subPath
: A subdirectory within the source folder where application code resides. Can be ignored if the source code resides at theroot
level.
-
Registry
source: registry: image: "" imagePullSecrets: - name: "" subPath: ""
registry
( Source code is an OCI image in a registry that contains application source)image
: Location of the source imageimagePullSecrets
: A list ofdockercfg
ordockerconfigjson
secret names required if the source image is private
subPath
: A subdirectory within the source folder where application code resides. Can be ignored if the source code resides at theroot
level.
The build
field on the image
resource can be used to configure env variables required during the build process, to configure resource limits on CPU
and memory
, and to configure pod tolerations, node selector, build timeout (specified in seconds), and affinity. To configure "Creation Time" of the built app image, pass in the unix EPOCH timestamp (i.e "1667243396") as a string or use "now" to use the current time.
build:
env:
- name: "name of env variable"
value: "value of the env variable"
resources:
requests:
cpu: "0.25"
memory: "128M"
limits:
cpu: "0.5"
memory: "256M"
tolerations:
- key: "key1"
operator: "Exists"
effect: "NoSchedule"
nodeSelector:
disktype: ssd
buildTimeout: 1600
creationTime: "now"
affinity:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/e2e-az-name
operator: In
values:
- e2e-az1
- e2e-az2
See the kubernetes documentation on setting environment variables and resource limits and requests for more information.
OCI images built by kpack can be signed with cosign when a cosign formatted secret is added to the service account configured on the image resource. The secret can be added using the cosign CLI or manually.
To create a cosign signing secret through the cosign CLI, when targetted to the Kubernetes cluster, use:
cosign generate-key-pair k8s://[NAMESPACE]/[NAME]
Alternatively, create the cosign secret and provide your own cosign key files manually to Kubernetes by running the following command:
% kubectl create secret generic <secret-name> --from-literal=cosign.password=<password> --from-file=</path/to/cosign.key>
<secret-name>
: The name of the secret. Ensure that the secret is created in the same namespace as the eventual image resource.<password>
: The password provided to encrypt the private key. If not present, an empty password will be used.</path/to/cosign.key>
: The cosign private key file generated withcosign generate-key-pair
.
After adding the cosign secret, the secret must be added to the list of secrets
on the service account that the image is configured with.
By default, the build number and build timestamp information will be added to the cosign signing annotations. Users can specify additional cosign annotations under the spec key.
cosign:
annotations:
- name: "annotationName"
value: "annotationValue"
One way these annotations can be viewed is through verifying cosign signatures. The annotations will be under the optional
key in the verified JSON response. For example, this can be done with:
% cosign verify -key /path/to/cosign.pub registry.example.com/project/image@sha256:<DIGEST>
Which provides a JSON response similar to:
{
"critical": {
"identity": {
"docker-reference": "registry.example.com/project/image"
}, "image": {
"docker-manifest-digest": "sha256:<DIGEST>"
}, "type": "cosign container image signature"
}, "optional": {
"buildNumber": "1",
"buildTimestamp": "20210827.175240",
"annotationName": "annotationValue"
}
}
Cosign signatures can be pushed to a different registry from where the built OCI image is written to. To enable this, add the corresponding annotation to the cosign secret resource.
metadata:
name: ...
namespace: ...
annotations:
kpack.io/cosign.repository: other.registry.com/project/image
data:
cosign.key: ...
cosign.password: ...
This will be equivalent to setting COSIGN_REPOSITORY
as specified in cosign Specifying Registry
The service account configured on the image resource must have registry credentials for the other registry configured in the secrets
list (they do not need to be configured as imagePullSecrets
).
It should be noted that if you wish to push the signatures to the same registry but a different repository as the image resource tag
, the credential used must have access to both paths. You cannot use two separate credentials for the same registry.
To sign docker images in a registry that does not fully support OCI media types, legacy equivalents can be used by adding the corresponding annotation to the cosign secret resource:
metadata:
name: ...
namespace: ...
annotations:
kpack.io/cosign.docker-media-types: "1"
data:
cosign.key: ...
cosign.password: ...
This will be equivalent to setting COSIGN_DOCKER_MEDIA_TYPES=1
as specified in the cosign registry-support
apiVersion: kpack.io/v1alpha2
kind: Image
metadata:
name: sample-image
namespace: build-namespace
spec:
tag: gcr.io/project-name/app
serviceAccountName: service-account
builder:
name: sample-builder
kind: ClusterBuilder
cache:
volume:
size: "1.5Gi" # Optional, if not set then the caching feature is disabled
storageClassName: "my-storage-class" # Optional, if not set then the default storageclass is used
failedBuildHistoryLimit: 5 # Optional, if not present defaults to 10
successBuildHistoryLimit: 5 # Optional, if not present defaults to 10
source:
git:
url: https://github.com/buildpack/sample-java-app.git
revision: main
build: # Optional
env:
- name: BP_JAVA_VERSION
value: 8.*
resources:
limits:
cpu: 100m
memory: 1G
requests:
cpu: 50m
memory: 512M
Source for github can also be specified in the ssh format if there is a corresponding ssh secret
source:
git:
url: [email protected]/buildpack/sample-java-app.git
revision: main
apiVersion: kpack.io/v1alpha2
kind: Image
metadata:
name: sample-image
namespace: build-namespace
spec:
tag: gcr.io/project-name/app
serviceAccountName: service-account
builder:
name: sample-builder
kind: ClusterBuilder
cache:
volume:
size: "1.5Gi" # Optional, if not set then the caching feature is disabled
storageClassName: "my-storage-class" # Optional, if not set then the default storageclass is used
failedBuildHistoryLimit: 5 # Optional, if not present defaults to 10
successBuildHistoryLimit: 5 # Optional, if not present defaults to 10
source:
blob:
url: https://storage.googleapis.com/build-service/sample-apps/spring-petclinic-2.1.0.BUILD-SNAPSHOT.jar
build: # Optional
env:
- name: BP_JAVA_VERSION
value: 8.*
resources:
limits:
cpu: 100m
memory: 1G
requests:
cpu: 50m
memory: 512M
When an image resource has successfully built with its current configuration, its status will report the up to date fully qualified built OCI image reference.
If you are using kubectl
this information is available with kubectl get <image-resource-name>
or kubectl describe <image-resource-name>
.
status:
conditions:
- lastTransitionTime: "2020-01-17T16:16:36Z"
status: "True"
type: Ready
latestImage: index.docker.io/sample/image@sha256:d3eb15a6fd25cb79039594294419de2328f14b443fa0546fa9e16f5214d61686
...
When a build fails its status will report the condition Succeeded=False.
status:
conditions:
- lastTransitionTime: "2020-01-17T16:13:48Z"
status: "False"
type: Ready
...
Notable deprecations from kpack.io/v1alpha1
include:
- Notary image signing
- Cloud Native Buildpack service bindings
kpack.io/v1alpha1
will eventually be removed entirely so please migrate existing Image resources to use kpack.io/v1alpha2
apis.
apiVersion
must be kpack.io/v1alpha1
The optional notary
field on the image
resource can be used to configure Notary image signing.
notary:
v1:
url: "https://example.com/notary"
secretRef:
name: "notary-secret"
v1.url
: The URL of the notary server.v1.secretRef.name
: A secret containing the encrypted private key and private key password.
To generate a signing key, use the following commands from the Docker Content Trust documentation:
% export DOCKER_CONTENT_TRUST_SERVER=<notary-server-url>
% docker trust key generate my-key
% docker trust signer add --key my-key.pub my-key registry.example.com/org/app
This will generate a private key in ~/.docker/trust/private
encrypted with the user provided password.
To create the notary secret used by kpack for image signing, run the following command:
% kubectl create secret generic <secret-name> --from-literal=password=<password> --from-file=$HOME/.docker/trust/private/<hash>.key
<secret-name>
: The name of the secret. Ensure that the secret is created in the same namespace as the eventual image resource.<password>
: The password provided to encrypt the private key.<hash>.key
: The private key file.