Chapter 3. Creating build inputs

A build input provides source content for builds to operate on. You can use the following build inputs to provide sources in OpenShift Container Platform, listed in order of precedence:

You can combine multiple inputs in a single build. However, as the inline Dockerfile takes precedence, it can overwrite any other file named Dockerfile provided by another input. Binary (local) input and Git repositories are mutually exclusive inputs.

You can use input secrets when you do not want certain resources or credentials used during a build to be available in the final application image produced by the build, or want to consume a value that is defined in a secret resource. External artifacts can be used to pull in additional files that are not available as one of the other build input types.

When you run a build:

The following example of a source definition includes multiple input types and an explanation of how they are combined. For more details on how each input type is defined, see the specific sections for each input type.

source:
  git:
    uri: https://github.com/openshift/ruby-hello-world.git 1
    ref: "master"
  images:
  - from:
      kind: ImageStreamTag
      name: myinputimage:latest
      namespace: mynamespace
    paths:
    - destinationDir: app/dir/injected/dir 2
      sourcePath: /usr/lib/somefile.jar
  contextDir: "app/dir" 3
  dockerfile: "FROM centos:7\nRUN yum install -y httpd" 4

When you supply a dockerfile value, the content of this field is written to disk as a file named dockerfile. This is done after other input sources are processed, so if the input source repository contains a Dockerfile in the root directory, it is overwritten with this content.

The source definition is part of the spec section in the BuildConfig:

source:
  dockerfile: "FROM centos:7\nRUN yum install -y httpd" 1

You can add additional files to the build process with images. Input images are referenced in the same way the From and To image targets are defined. This means both container images and image stream tags can be referenced. In conjunction with the image, you must provide one or more path pairs to indicate the path of the files or directories to copy the image and the destination to place them in the build context.

The source path can be any absolute path within the image specified. The destination must be a relative directory path. At build time, the image is loaded and the indicated files and directories are copied into the context directory of the build process. This is the same directory into which the source repository content is cloned. If the source path ends in /. then the content of the directory is copied, but the directory itself is not created at the destination.

Image inputs are specified in the source definition of the BuildConfig:

source:
  git:
    uri: https://github.com/openshift/ruby-hello-world.git
    ref: "master"
  images: 1
  - from: 2
      kind: ImageStreamTag
      name: myinputimage:latest
      namespace: mynamespace
    paths: 3
    - destinationDir: injected/dir 4
      sourcePath: /usr/lib/somefile.jar 5
  - from:
      kind: ImageStreamTag
      name: myotherinputimage:latest
      namespace: myothernamespace
    pullSecret: mysecret 6
    paths:
    - destinationDir: injected/dir
      sourcePath: /usr/lib/somefile.jar

$ oc secrets link builder dockerhub

When specified, source code is fetched from the supplied location.

If you supply an inline Dockerfile, it overwrites the Dockerfile in the contextDir of the Git repository.

The source definition is part of the spec section in the BuildConfig:

source:
  git: 1
    uri: "https://github.com/openshift/ruby-hello-world"
    ref: "master"
  contextDir: "app/dir" 2
  dockerfile: "FROM openshift/ruby-22-centos7\nUSER example" 3

If the ref field denotes a pull request, the system uses a git fetch operation and then checkout FETCH_HEAD.

When no ref value is provided, OpenShift Container Platform performs a shallow clone (--depth=1). In this case, only the files associated with the most recent commit on the default branch (typically master) are downloaded. This results in repositories downloading faster, but without the full commit history. To perform a full git clone of the default branch of a specified repository, set ref to the name of the default branch (for example main).

source:
  git:
    uri: "https://github.com/openshift/ruby-hello-world"
    ref: "master"
    httpProxy: http://proxy.example.com
    httpsProxy: https://proxy.example.com
    noProxy: somedomain.com, otherdomain.com

$ oc annotate secret mysecret \
    'build.openshift.io/source-secret-match-uri-1=ssh://bitbucket.atlassian.com:7999/*'

kind: Secret
apiVersion: v1
metadata:
  name: matches-all-corporate-servers-https-only
  annotations:
    build.openshift.io/source-secret-match-uri-1: https://*.mycorp.com/*
data:
  ...
---
kind: Secret
apiVersion: v1
metadata:
  name: override-for-my-dev-servers-https-only
  annotations:
    build.openshift.io/source-secret-match-uri-1: https://mydev1.mycorp.com/*
    build.openshift.io/source-secret-match-uri-2: https://mydev2.mycorp.com/*
data:
  ...

apiVersion: "build.openshift.io/v1"
kind: "BuildConfig"
metadata:
  name: "sample-build"
spec:
  output:
    to:
      kind: "ImageStreamTag"
      name: "sample-image:latest"
  source:
    git:
      uri: "https://github.com/user/app.git"
    sourceSecret:
      name: "basicsecret"
  strategy:
    sourceStrategy:
      from:
        kind: "ImageStreamTag"
        name: "python-33-centos7:latest"

$ oc create secret generic <secret_name> --from-file=<path/to/.gitconfig>

[http]
        sslVerify=false

Streaming content from a local file system to the builder is called a Binary type build. The corresponding value of BuildConfig.spec.source.type is Binary for these builds.

This source type is unique in that it is leveraged solely based on your use of the oc start-build.

To utilize binary builds, invoke oc start-build with one of these options:

In each of the previously listed cases:

Instead of a file name, you can pass a URL with HTTP or HTTPS schema to --from-file and --from-archive. When using --from-file with a URL, the name of the file in the builder image is determined by the Content-Disposition header sent by the web server, or the last component of the URL path if the header is not present. No form of authentication is supported and it is not possible to use custom TLS certificate or disable certificate validation.

When using oc new-build --binary=true, the command ensures that the restrictions associated with binary builds are enforced. The resulting BuildConfig has a source type of Binary, meaning that the only valid way to run a build for this BuildConfig is to use oc start-build with one of the --from options to provide the requisite binary data.

The Dockerfile and contextDir source options have special meaning with binary builds.

Dockerfile can be used with any binary build source. If Dockerfile is used and the binary stream is an archive, its contents serve as a replacement Dockerfile to any Dockerfile in the archive. If Dockerfile is used with the --from-file argument, and the file argument is named Dockerfile, the value from Dockerfile replaces the value from the binary stream.

In the case of the binary stream encapsulating extracted archive content, the value of the contextDir field is interpreted as a subdirectory within the archive, and, if valid, the builder changes into that subdirectory before executing the build.

In some scenarios, build operations require credentials or other configuration data to access dependent resources, but it is undesirable for that information to be placed in source control. You can define input secrets and input config maps for this purpose.

For example, when building a Java application with Maven, you can set up a private mirror of Maven Central or JCenter that is accessed by private keys. To download libraries from that private mirror, you have to supply the following:

For security reasons, you do not want to expose your credentials in the application image.

This example describes a Java application, but you can use the same approach for adding SSL certificates into the /etc/ssl/certs directory, API keys or tokens, license files, and more.

apiVersion: v1
kind: Secret
metadata:
  name: test-secret
  namespace: my-namespace
type: Opaque 1
data: 2
  username: <username> 3
  password: <password>
stringData: 4
  hostname: myapp.mydomain.com 5

FROM centos/ruby-22-centos7

USER root
COPY ./secret-dir /secrets
COPY ./config /

# Create a shell script that will output secrets and ConfigMaps when the image is run
RUN echo '#!/bin/sh' > /input_report.sh
RUN echo '(test -f /secrets/secret1 && echo -n "secret1=" && cat /secrets/secret1)' >> /input_report.sh
RUN echo '(test -f /config && echo -n "relative-configMap=" && cat /config)' >> /input_report.sh
RUN chmod 755 /input_report.sh

CMD ["/bin/sh", "-c", "/input_report.sh"]

It is not recommended to store binary files in a source repository. Therefore, you must define a build which pulls additional files, such as Java .jar dependencies, during the build process. How this is done depends on the build strategy you are using.

For a Source build strategy, you must put appropriate shell commands into the assemble script:

#!/bin/sh
APP_VERSION=1.0
wget http://repository.example.com/app/app-$APP_VERSION.jar -O app.jar

#!/bin/sh
exec java -jar app.jar

For a Docker build strategy, you must modify the Dockerfile and invoke shell commands with the RUN instruction:

FROM jboss/base-jdk:8

ENV APP_VERSION 1.0
RUN wget http://repository.example.com/app/app-$APP_VERSION.jar -O app.jar

EXPOSE 8080
CMD [ "java", "-jar", "app.jar" ]

In practice, you may want to use an environment variable for the file location so that the specific file to be downloaded can be customized using an environment variable defined on the BuildConfig, rather than updating the Dockerfile or assemble script.

You can choose between different methods of defining environment variables:

You can supply builds with a .docker/config.json file with valid credentials for private container registries. This allows you to push the output image into a private container image registry or pull a builder image from the private container image registry that requires authentication.

You can supply credentials for multiple repositories within the same registry, each with credentials specific to that registry path.

The .docker/config.json file is found in your home directory by default and has the following format:

auths:
  index.docker.io/v1/: 1
    auth: "YWRfbGzhcGU6R2labnRib21ifTE=" 2
    email: "user@example.com" 3
  docker.io/my-namespace/my-user/my-image: 4
    auth: "GzhYWRGU6R2fbclabnRgbkSp=""
    email: "user@example.com"
  docker.io/my-namespace: 5
    auth: "GzhYWRGU6R2deesfrRgbkSp=""
    email: "user@example.com"

You can define multiple container image registries or define multiple repositories in the same registry. Alternatively, you can also add authentication entries to this file by running the docker login command. The file will be created if it does not exist.

Kubernetes provides Secret objects, which can be used to store configuration and passwords.

As with pod environment variables, build environment variables can be defined in terms of references to other resources or variables using the Downward API. There are some exceptions, which are noted.

You can also manage environment variables defined in the BuildConfig with the oc set env command.

Service serving certificate secrets are intended to support complex middleware applications that need out-of-the-box certificates. It has the same settings as the server certificates generated by the administrator tooling for nodes and masters.

Other pods can trust cluster-created certificates, which are only signed for internal DNS names, by using the certificate authority (CA) bundle in the /var/run/secrets/kubernetes.io/serviceaccount/service-ca.crt file that is automatically mounted in their pod.

The signature algorithm for this feature is x509.SHA256WithRSA. To manually rotate, delete the generated secret. A new certificate is created.

To use a secret, a pod needs to reference the secret. A secret can be used with a pod in three ways:

Volume type secrets write data into the container as a file using the volume mechanism. imagePullSecrets use service accounts for the automatic injection of the secret into all pods in a namespaces.

When a template contains a secret definition, the only way for the template to use the provided secret is to ensure that the secret volume sources are validated and that the specified object reference actually points to an object of type Secret. Therefore, a secret needs to be created before any pods that depend on it. The most effective way to ensure this is to have it get injected automatically through the use of a service account.

Secret API objects reside in a namespace. They can only be referenced by pods in that same namespace.

Individual secrets are limited to 1MB in size. This is to discourage the creation of large secrets that would exhaust apiserver and kubelet memory. However, creation of a number of smaller secrets could also exhaust memory.