docs/secrets.md at ff746c93247d7890dc148fae3771c9b12bafbff0

6.8 KiB

Raw Blame History

title

linkTitle

weight

description

keywords

Secret mounts

Secret mounts expose secrets to the build containers, as files or environment variables. You can use secret mounts to pass sensitive information to your builds, such as API tokens, passwords, or SSH keys. You mount the secrets to the RUN instructions that need to access them, similar to how you would define a bind mount or cache mount.

Passing secrets

To pass a secret to a build, use the docker build --secret flag, or the equivalent options for Bake.

$ docker build --secret id=mytoken,src=$HOME/.aws/credentials .

variable "HOME" {
  default = null
}

target "default" {
  secret = [
    "id=mytoken,src=${HOME}/.aws/credentials"
  ]
}

Sources

The source of a secret can be either a file or an environment variable. When you use the CLI or Bake, the type can be detected automatically. You can also specify it explicitly with type=file or type=env.

The following example mounts the environment variable KUBECONFIG to secret ID kube, as a file in the build container at /run/secrets/kube.

$ docker build --secret id=kube,env=KUBECONFIG .

When you use secrets from environment variables, you can omit the env parameter to bind the secret to a file with the same name as the variable. In the following example, the value of the API_TOKEN variable is mounted to /run/secrets/API_TOKEN in the build container.

$ docker build --secret id=API_TOKEN .

Target

By default, secrets are mounted as files located at /run/secrets/<id>. You can customize how the secrets get mounted in the build container using the target and env options for the RUN --mount flag in the Dockerfile.

The following example takes secret id aws and mounts it to /run/secrets/aws in the build container.

RUN --mount=type=secret,id=aws \
    AWS_SHARED_CREDENTIALS_FILE=/run/secrets/aws \
    aws s3 cp ...

To mount a secret as a file with a different name, use the target option in the --mount flag.

RUN --mount=type=secret,id=aws,target=/root/.aws/credentials \
    aws s3 cp ...

To mount a secret as an environment variable instead of a file, use the env option in the --mount flag.

RUN --mount=type=secret,id=aws-key-id,env=AWS_ACCESS_KEY_ID \
    --mount=type=secret,id=aws-secret-key,env=AWS_SECRET_ACCESS_KEY \
    --mount=type=secret,id=aws-session-token,env=AWS_SESSION_TOKEN \
    aws s3 cp ...

It's possible to use the target and env options together to mount a secret as both a file and an environment variable.

SSH mounts

If the credential you want to use in your build is an SSH agent socket or key, you can use the SSH mount instead of a secret mount. Cloning private Git repositories is a common use case for SSH mounts.

The following example clones a private GitHub repository using a Dockerfile SSH mount.

# syntax=docker/dockerfile:1
FROM alpine
ADD git@github.com:me/myprivaterepo.git /src/

To pass an SSH socket the build, you use the docker build --ssh flag, or equivalent options for Bake.

$ docker buildx build --ssh default .

Git authentication for remote contexts

BuildKit supports two pre-defined build secrets, GIT_AUTH_TOKEN and GIT_AUTH_HEADER. Use them to specify HTTP authentication parameters when building with remote, private Git repositories, including:

Building with a private Git repository as build context
Fetching private Git repositories in a build with ADD

For example, say you have a private GitLab project at https://gitlab.com/example/todo-app.git, and you want to run a build using that repository as the build context. An unauthenticated docker build command fails because the builder isn't authorized to pull the repository:

$ docker build https://gitlab.com/example/todo-app.git
[+] Building 0.4s (1/1) FINISHED
 => ERROR [internal] load git source https://gitlab.com/example/todo-app.git
------
 > [internal] load git source https://gitlab.com/example/todo-app.git:
0.313 fatal: could not read Username for 'https://gitlab.com': terminal prompts disabled
------

To authenticate the builder to the Git server, set the GIT_AUTH_TOKEN environment variable to contain a valid GitLab access token, and pass it as a secret to the build:

$ GIT_AUTH_TOKEN=$(cat gitlab-token.txt) docker build \
  --secret id=GIT_AUTH_TOKEN \
  https://gitlab.com/example/todo-app.git

The GIT_AUTH_TOKEN also works with ADD to fetch private Git repositories as part of your build:

FROM alpine
ADD https://gitlab.com/example/todo-app.git /src

HTTP authentication scheme

By default, Git authentication over HTTP uses the Bearer authentication scheme:

Authorization: Bearer <GIT_AUTH_TOKEN>

If you need to use a Basic scheme, with a username and password, you can set the GIT_AUTH_HEADER build secret:

$ export GIT_AUTH_TOKEN=$(cat gitlab-token.txt)
$ export GIT_AUTH_HEADER=basic
$ docker build \
  --secret id=GIT_AUTH_TOKEN \
  --secret id=GIT_AUTH_HEADER \
  https://gitlab.com/example/todo-app.git

BuildKit currently only supports the Bearer and Basic schemes.

Multiple hosts

You can set the GIT_AUTH_TOKEN and GIT_AUTH_HEADER secrets on a per-host basis, which lets you use different authentication parameters for different hostnames. To specify a hostname, append the hostname as a suffix to the secret ID:

$ export GITLAB_TOKEN=$(cat gitlab-token.txt)
$ export GERRIT_TOKEN=$(cat gerrit-username-password.txt)
$ export GERRIT_SCHEME=basic
$ docker build \
  --secret id=GIT_AUTH_TOKEN.gitlab.com,env=GITLAB_TOKEN \
  --secret id=GIT_AUTH_TOKEN.gerrit.internal.example,env=GERRIT_TOKEN \
  --secret id=GIT_AUTH_HEADER.gerrit.internal.example,env=GERRIT_SCHEME \
  https://gitlab.com/example/todo-app.git

6.8 KiB Raw Blame History