We are using the Google Container Registry to store our Docker images.
To authorize our build instances we place long-lived access tokens in .docker/config.json as described in the docs.
This works perfectly fine until someone (i.e. some Makefile) uses gcloud docker -- push ... to push to the registry (instead of e.g. docker push ...). gcloud will replace the existing, long-lived credentials with short-lived ones that expire after some time. Thus subsequent builds may fail, depending on the exact timing.
My Question: How can I prevent gcloud docker ... from messing with my provisioned credentials?
I've tried chattr +i .docker/config.json, but this just makes gcloud complain.
From https://cloud.google.com/sdk/gcloud/reference/docker:
The gcloud docker command group wraps docker commands, so that gcloud can inject the appropriate fresh authentication token into requests that interact with the docker registry.
The only thing that gcloud docker does is change these credentials, then invoke the docker CLI. If you don't want it to change the credentials, there's no reason not to just call docker directly.
One workaround might be to use an alternate configuration file location for your long-lived credentials; per https://docs.docker.com/engine/reference/commandline/cli/:
Options:
--config string Location of client config files (default "/root/.docker")
Related
I was wondering is it possible to use cached docker images in gitlab registry for gitlab-ci?
for example, I want to use node:16.3.0-alpine docker image, can I cache it in my gitlab registry and pull it from that and speed up my gitlab ci instead of pulling it from docker hub?
Yes, GitLab's dependency proxy features allow you to configure GitLab as a "pull through cache". This is also beneficial for working around rate limits of upstream sources like dockerhub.
It should be faster in most cases to use the dependency proxy, but not necessarily so. It's possible that dockerhub can be more performant than a small self-hosted server, for example. GitLab runners are also remote with respect to the registry and not necessarily any "closer" to the GitLab registry than any other registry over the internet. So, keep that in mind.
As a side note, the absolute fastest way to retrieve cached images is to self-host your GitLab runners and hold images directly on the host. That way, when jobs start, if the image already exists on the host, the job will start immediately because it does not need to pull the image (depending on your pull configuration). (that is, assuming you're using images in the image: declaration for your job)
I'm using a corporate Gitlab instance where for some reason the Dependency Proxy feature has been disabled. The other option you have is to create a new Docker image on your local machine, then push it into the Container Registry of your personal Gitlab project.
# First create a one-line Dockerfile containing "FROM node:16.3.0-alpine"
docker pull node:16.3.0-alpine
docker build . -t registry.example.com/group/project/image
docker login registry.example.com -u <username> -p <token>
docker push registry.example.com/group/project/image
where the image tag should be constructed based on the example given on your project's private Container Registry page.
Now in your CI job, you just change image: node:16.3.0-alpine to image: registry.example.com/group/project/image. You may have to run the docker login command (using a deploy token for credentials, see Settings -> Repository) in the before_script section -- I think maybe newer versions of Gitlab will have the runner authenticate to the private Container Registry using system credentials, but that could vary depending on how it's configured.
In order to deploy an application using docker and a remote registry:
Using docker client, execute docker login so that the credentials will be stored either directly in $HOME/.docker/config.json or on a credential store specified also in $HOME/.docker/config.json. Then use the docker create command to start the application.
In kubernetes, a secret can be generated using the docker registry username and password. Then, the secret can be injected in the helm-chart using imagePullSecret. then, helm install command can instruct kubelet to pull the image inside the created container inside the scheduled pod. To update the image registry, the image name and pull secret can be updated before re-installation.
I have three questions:
How can I set the username and password or inject these credentials for the services in docker-compose without having to run docker login first in each deployment host ? (as in nu
Can I populate a credential store specified in a $HOME/.docker/config.json using docker login command on one machine, then specify the same credential store in $HOME/.docker/config.json of another machine, then use the answer of the previous question to inject or pull the credentials
if the docker daemon checks for the credentials inside the credential stores that is specified in $HOME/.docker/config.json, then what is the use of the helper program ?
Since Docker Hub only allows 1 private repo, I wonder if there is any way to use Github or Gitlab, etc., to download the images? for instance:
FROM git#github.com/username/repo
...
...
...
Very easy with an account on gitlab.com. GitLab provides a Docker registry linked to projects and you can have unlimited private projects:
Create a project my-docker-project
Go to Package and Registries > Container registries, you should see a few commands to access your registry
Connect your machine to this registry using a command like:
# Will prompt for login/pass
docker login registry.gitlab.com
You'll need an access token or deploy token with read_registry and write_registry scopes. You can generate one via your profile Preferences > Access token. Login is the token name and password the secret token provided.
You can now push Docker images with commands such as:
# Push an image
docker push registry.gitlab.com/YourUsernameOrGroup/my-docker-project
# Push an image on a sub-path
docker push registry.gitlab.com/YourUsernameOrGroup/my-docker-project/myimage
You can then use the image in a Dockerfile by referencing its URL such as:
FROM registry.gitlab.com/YourUsernameOrGroup/my-docker-project
# ...
Of course the machine from which you build must be authenticated on related GitLab registry using docker login command above (or the project must be public)
both have excellent package registry services
For GitHub GPR
For GitLab GCR
Both have excellent features like use it directly from Dockerfile as you want for example.
I have a public example, you can check it in Github with node.js which uses GPR to store the build image/package.
I'm building an application that I will run in GKE. This application will use shell commands (for now) to build docker images and try to push them to GCR. I'm finding that when I try to do this from a pod running in GKE I get authentication problems. I'm having trouble figuring out why these authentication problems are happening.
Here's a list of all of the debugging I've done so far. At the highest level, my GKE clusters have the https://www.googleapis.com/auth/devstorage.read_write oauth scope. When I examine the permissions on the underlying GCE instance, I see these permissions - note the Read Write value for Storage:
Now, when I SSH into that instance using the console and list the docker images I see the image used by GKE when spinning up my pod:
paymahn#gke-prod-478557c-default-pool-e9314f46-d9mn ~ $ docker image ls
REPOSITORY TAG IMAGE ID CREATED SIZE
gcr.io/gadic-310112/server latest 8f8a22237c31 2 days ago 1.85GB
...
However, if I try to manually pull that image while SSH-ed into the GCP instance, I get an authentication problem:
paymahn#gke-prod-478557c-default-pool-e9314f46-d9mn ~ $ docker pull gcr.io/gadic-310112/server:latest
Error response from daemon: unauthorized: You don't have the needed permissions to perform this operation, and you may have invalid credentials. To authenticate your request, follow the steps in: https://cloud.google.com/container-registry/docs/advanced-authentication
I also looked at the service account 65106360748-compute#developer.gserviceaccount.com which is the default compute instance service account. Here are the permissions it has (I manually added the Storage Object Creator role):
Adding the Storage Object Creator role to that service account didn't help.
Is my approach to authentication here fundamentally flawed? It seems like I have all the right pieces in place to pull/push from GCR from GKE. Maybe there's an extra step I need to do for the docker client to authenticate?
Figured it out. I had to:
make a service account with the roles/storage.objectAdmin
generate a key for that service account
store that key as a secret in GKE
Mount that secret into my pods
run gcloud auth activate-service-account --key-file <path to key>
run gcloud auth configure-docker
Once all of that was done, my pods could pull from and push to GCR.
I have a private GitLab project with a pipeline for building and pushing a Docker image. Therefore I have to authenticate to GitLab's Docker registry first.
Research
I read Authenticating to the Container Registry with GitLab CI/CD:
There are three ways to authenticate to the Container Registry via GitLab CI/CD which depend on the visibility of your project.
Available for all projects, though more suitable for public ones:
Using the special CI_REGISTRY_USER variable: The user specified by this variable is created for you in order to push to the Registry connected to your project. Its password is automatically set with the CI_REGISTRY_PASSWORD variable. This allows you to automate building and deploying your Docker images and has read/write access to the Registry. This is ephemeral, so it’s only valid for one job. You can use the following example as-is:
docker login -u $CI_REGISTRY_USER -p $CI_REGISTRY_PASSWORD $CI_REGISTRY
For private and internal projects:
Using a personal access token: You can create and use a personal access token in case your project is private:
For read (pull) access, the scope should be read_registry.
For read/write (pull/push) access, use api.
Replace the <username> and <access_token> in the following example:
docker login -u <username> -p <access_token> $CI_REGISTRY
Using the GitLab Deploy Token: You can create and use a special deploy token with your private projects. It provides read-only (pull) access to the Registry. Once created, you can use the special environment variables, and GitLab CI/CD will fill them in for you. You can use the following example as-is:
docker login -u $CI_DEPLOY_USER -p $CI_DEPLOY_PASSWORD $CI_REGISTRY
and Container Registry:
With the update permission model we also extended the support for accessing Container Registries for private projects.
Version history
Your jobs can access all container images that you would normally have access to. The only implication is that you can push to the Container Registry of the project for which the job is triggered.
This is how an example usage can look like:
test:
script:
- docker login -u gitlab-ci-token -p $CI_JOB_TOKEN $CI_REGISTRY
- docker pull $CI_REGISTRY/group/other-project:latest
- docker run $CI_REGISTRY/group/other-project:latest
I tried the first and the fourth way and I could authenticate.
Question
What are the pros and cons? I guess the third way is for deployment only, not for building and pushing. Same could be for the second way. Is that right?
And why is the fourth way not listed in the other documentation? Is that way deprecated?
I prefer the fourth option. A note: "If a user creates one named gitlab-deploy-token, the username and token of the deploy token is automatically exposed to the CI/CD jobs as CI/CD variables: CI_DEPLOY_USER and CI_DEPLOY_PASSWORD respectively.
When creating deploy token, you can grant permission read/write to registry/package registry.
The CI_REGISTRY_PASSWORD is ephemeral so avoid using it if you have multiple deploy jobs (which need to pull private image) run parallel.
I believe the differences are just about user skill and permissions.
The first way anyone can do since the variables are automatically present in a running job.
Second, anyone, with any permissions, can create a personal access token (but has an extra step compared to 1 to create the access token).
Third, someone with the correct permissions could create a deploy key. Deploy keys don't give access to the API like personal access tokens can, and only have permission to pull/read the data in the repository, they cannot write/push.
Fourth option, it allows you to both read/pull container images from the registry, but it also allows you to push to the registry. This is helpful if you have a CI step that builds an app in an image, or anything else where you're generating a container image and want to push it into the registry (so another step in the pipeline can pull it down and use it). My guess is that this option isn't listed with the others since it's meant for the building of container images. You probably could use it like any of the others though.