Image Building
Images are configured within the images
section of the devspace.yaml
.
Examples
- Docker Hub
- Custom Registry
- Custom Tagging
- kaniko
- Build Command
- Disable
images:
backend:
image: my-docker-username/appbackend
database:
image: my-docker-company/appbackend
images:
backend:
image: image.company.tld/appbackend
database:
image: image.company.tld/appbackend
images:
backend:
image: image.company.tld/appbackend
tags:
- latest
- dev-${DEVSPACE_GIT_COMMIT}-${DEVSPACE_RANDOM}
database:
image: image.company.tld/appbackend
images:
backend:
image: john/appbackend
build:
kaniko:
cache: true
Kaniko is an open-source tool that lets you build images inside containers that run on top of Kubernetes. DevSpace uses kaniko as automatic fallback if Docker is not installed. To force image building using kaniko, set build.kaniko.cache: true
as shown above. Learn more about configuring kaniko builds.
images:
backend:
image: john/appbackend
build:
custom:
command: ./build
args:
- "--arg1"
- "arg-value-1"
- "--arg2"
- "arg-value-2"
images:
frontend:
image: dscr.io/${DEVSPACE_USERNAME}/appfrontend
build:
disabled: true
backend:
image: john/appbackend
To speed up the build process, the images you specify under images
will all be built in parallel (unless you use the --build-sequential
flag).
If you specify multiple build tools, DevSpace will try to use them in the following order:
disabled
custom
docker
(uses kaniko as fallback if Docker host not reachable)kaniko
Run Image Building
When you run one of the following commands, DevSpace will run the image building process:
devspace build
(only image building without deployment)devspace deploy
(before deploying the application)devspace dev
(before deploying the application and starting the development mode)
Important Flags
The following flags are available for all commands that trigger image building:
--build-sequential
build images sequentially instead of in parallel (provides more detailed logs for each image)-b / --force-build
rebuild all images (even if they could be skipped because context and Dockerfile have not changed since the latest build)
Image Building Process
DevSpace loads the images
configuration from devspace.yaml
and builds all images in parallel. The multi-threaded, parallel build process of DevSpace speeds up image building drastically, especially when building many images and using remote build methods.
1. Load Dockerfile
DevSpace loads the contents of the Dockerfile specified in image[*].dockerfile
(defaults to ./Dockerfile
).
Dockerfile paths used in dockerfile
should be relative to the devspace.yaml
.
2. Override Entrypoint (if needed)
DevSpace allows you to apply an in-memory override of a Dockerfile's ENTRYPOINT
by configuring the entrypoint
option for the image. Similar to the Dockerfile ENTRYPOINT
, the images[*].entrypoint
option should be defined as an array.
Configuring ENTRYPOINT
overrides can be particularly useful when defining different config profiles in your devspace.yaml
.
3. Load Build Context
DevSpace loads the context to build this image as specified in context
(defaults to ./
). The context path serves as root directory for Dockerfile statements like ADD
or COPY
.
Context paths used in context
should be relative to the devspace.yaml
.
What does "context" mean in terms of image building?
The context is archived and sent to the Docker daemon before starting to process the Dockerfile. All references of local files within the Dockerfile are relative to the root directory of the context.
That means that a Dockerfile statement such as COPY ./src /app
would copy the folder src/
within the context path into the path /app
within the container image. So, if the context would be /my/project/database
, for example, the folder that would be copied into /app
would have the absolute path /my/project/database/src
on your local computer.
4. Skip Build & Push (if possible)
DevSpace tries to speed up image building by skipping images when they have not changed since the last build process. To do this, DevSpace caches the following information after building an image:
- a hash of the
Dockerfile
used to build the image (including ENTRYPOINT override) - a hash over all files in the
context
used to build this image (excluding files in.dockerignore
)
The next time you trigger the image building process, DevSpace will generate these hashes again and compare them to the hashes of the last image building process. If all newly generated hashes are equal to the old ones stored during the last image building process, then nothing has changed and DevSpace will skip this image.
You can use the -b / --force-build
flag to tell DevSpace to build all images even if nothing has changed.
5. Build Image
DevSpace uses one of the following build tools to create an image based on your Dockerfile and the provided context:
docker
for building images using a Docker daemon (default, prefers Docker daemon of local Kubernetes clusters)kaniko
for building images directly inside Kubernetes (automatic fallback fordocker
)custom
for building images with a custom build command (e.g. for using Google Cloud Build)disabled
if this image should not be built (especially useful for configprofiles
)
6. Tag Image
DevSpace automatically tags all images after building them using a tagging schema that you can customize using the tag
option. By default, this option is configured to generate a random string consisting of 5 characters.
Learn more about defining a custom tagging schema.
Before deploying your application, DevSpace will use the newly generated image tags and replace them in-memory in the values for your Helm charts and Kubernetes manifests, so they will be deployed using the newest image tag.
7. Push Image
DevSpace automatically pushes your images to the respective registry that should be specified as part of the image
option. Just as with regular Docker images, DevSpace uses Docker Hub if no registry hostname is provided within image
.
You can skip this step when proving the --skip-push
flag. Beware that deploying your application after using --skip-push
will only work when your images have already been pushed to the registry or when you are using a local Kubernetes cluster (e.g. minikube).
8. Create Pull Secrets
When deploying your application via DevSpace, Kubernetes needs to be able to pull your images from the registry that is used to store your images when pushing them. For this purpose, Kubernetes relies on so-called image pull secrets. DevSpace can automatically create these secrets for you, if you configure createPullSecret: true
for the respective image in your devspace.yaml
.
You do not need to change anything in your Kubernetes manifests or Helm charts to use the image pull secrets that DevSpace creates because DevSpace adds these secrets to the default service account in the namespace used to deploy your project.
After running devspace build
, devspace deploy
or devspace dev
, you should be able to see the auto-generated pull secrets created by DevSpace when you run the following command:
kubectl get serviceaccount default -o yaml
Take a look at the imagePullSecrets
section of the output showing the yaml definition of the service account default
.
Advanced Topics
Authentication (Registry)
DevSpace uses the same credential store as Docker. So, if you already have Docker installed and you have logged in to a private registry before, DevSpace will be able to push to this registry. So, if you want to push to a registry using DevSpace, simply authenticate using this command:
docker login # for Docker Hub
docker login [REGISTRY] # for any other registry (e.g. my-registry.tld)
If you do not have Docker installed, DevSpace initializes a Docker credential store for you and stores your login information just like Docker would do it.
Skip Push (Local Clusters)
If you are using a local Kubernetes cluster, DevSpace will try to build the image using the Docker deamon of this local cluster. If this process is successful, DevSpace will skip the step of pushing the image to a registry as it is not required for deploying your application.