I don't maintain this repository anymore. Feel free to fork if needed.
This application creates a docker config.json (as a Kubernetes secret) that can authenticate docker clients to Amazon ECR. It is using the ECR GetAuthorizationToken API to fetch the token from an Amazon Region.
A docker config.json file looks like this, which may be found in $HOME/.docker/config.json
after using docker login
.
{
"auths": {
"https://${AWS_PROFILE}.dkr.ecr.us-east-1.amazonaws.com": {
"auth": "....."
}
}
}
This application is like a running cron job that does aws ecr get-login
, creates a docker config.json file, then create Kubernetes secret out of it.
WARNING!! If you need to run this in production environments, please build your own Docker image by following the How To Build this Project step.
-
Deploy this application as a Pod in the namespace to create the Secret.
-
This Pod authenticates to AWS to get a new ECR docker login token according to the
-region
flag. -
It then creates a config.json according to the values retrieved from AWS.
# config.json template const cfgTemplate = `{ "auths": { "{{ .registry }}": { "auth": "{{ .token }}" } } }`
-
It then creates or updates the Secret according to the
-secretName
flag in the current namespace.apiVersion: v1 type: Secret metadata: name: xxxx # -secretName namespace: xxxx # current pod namespace data: config.json: xxxxx # base64 encoded
-
You can specify secret type by using
-secretType
flag. This can be useful, when running kubernetes nodes outside of AWS, as in this scenario kubelet does not have pull access to ECR. Specifying-secretType=kubernetes.io/dockerconfigjson
will automate creation of a secret, that can be used in manifests requiringimagePullSecrets
apiVersion: v1 data: .dockerconfigjson: xxxxx kind: Secret type: kubernetes.io/dockerconfigjson metadata: name: xxxx namespace: xxxx
-
It repeats the process according to the specified
-interval
flag. Refreshing the config.json file content in the Secret over the time specified. -
You can now use this kubernetes secret and mount it to any pod that has a docker client that authenticates to ECR.
We are using Weave Flux to operate our GitOps deployment process. The Weave Flux operator currently does not support authentication to ECR (issue #539). As a workaround, we can use the --docker-config
flag and mount a custom config.json
in the flux Pod (issue #1065).
The problem is ECR token expires every 12 hours, and we need to find a way to ensure that the config.json authentication token is rotated in an automated and secure way.
Complete the Deployment Guide, noting the secret name and then follow Flux Guide.
- IAM Role Requirement
- Deployment
- Check the ECR Secret Creator Pod's logs
- Check the Created Secret
- How to Build this Project
If you are NOT using kube2iam, skip to the Deployment Step and ensure that your Pod's EC2 instance can authenticate to your AWS Account via AWS API Keys, or AWS EC2 IAM Role.
Create IAM Policy
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": "ecr:GetAuthorizationToken",
"Resource": "*"
}
]
}
aws iam create-policy --policy-name ${GET_ECR_AUTH_IAM_POLICY} --policy-document file://iam-policy.json --description "A policy that can get ECR authorization token"
Create IAM Role
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Principal": {
"Service": "ec2.amazonaws.com"
},
"Action": "sts:AssumeRole"
},
{
"Sid": "",
"Effect": "Allow",
"Principal": {
"AWS": "arn:aws:iam::${AWS_PROFILE}:role/${IAM_K8S_NODE_ROLE}"
},
"Action": "sts:AssumeRole"
}
]
}
aws iam create-role --role-name ${GET_ECR_AUTH_IAM_ROLE} \
--assume-role-policy-document file://ec2-iam-trust.json
Attach the IAM Policy
aws iam attach-role-policy --role-name ${GET_ECR_AUTH_IAM_ROLE} \
--policy-arn arn:aws:iam::${AWS_PROFILE}:policy/${GET_ECR_AUTH_IAM_POLICY}
Create the following RBAC and deploy resources yaml files and run kubectl apply -f
.
---
kind: ServiceAccount
apiVersion: v1
metadata:
name: ecr-k8s-secret-creator
namespace: ${SECRET_NAMESPACE}
---
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: ecr-k8s-secret-creator
rules:
- apiGroups: [""]
resources: ["secrets"]
verbs: ["get", "list", "watch", "create", "update", "patch", "delete"]
---
kind: ClusterRoleBinding
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: ecr-k8s-secret-creator
roleRef:
kind: ClusterRole
name: ecr-k8s-secret-creator
apiGroup: rbac.authorization.k8s.io
subjects:
- kind: ServiceAccount
name: ecr-k8s-secret-creator
namespace: ${SECRET_NAMESPACE}
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
name: ecr-k8s-secret-creator
namespace: ${SECRET_NAMESPACE}
spec:
replicas: 1
selector:
matchLabels:
app: ecr-k8s-secret-creator
template:
metadata:
labels:
app: ecr-k8s-secret-creator
# if using kube2iam
# annotations:
# iam.amazonaws.com/role: arn:aws:iam::${AWS_PROFILE}:role/${GET_ECR_AUTH_IAM_ROLE}
spec:
serviceAccount: ecr-k8s-secret-creator
containers:
- image: bzon/ecr-k8s-secret-creator:latest
name: ecr-k8s-secret-creator
args:
- "-secretName=ecr-docker-secret"
- "-region=us-east-1"
# the default profile is the AWS account where the kubernetes cluster is running
# - "-profile=${AWS_ACCOUNT_ID}"
# the default interval for fetching new ecr token is 200 seconds
# - "-interval=300"
resources:
requests:
cpu: 5m
memory: 16Mi
limits:
cpu: 20m
memory: 32Mi
{"level":"info","msg":"appVersion: 0.1.0","time":"2018-09-29T16:48:17Z"}
{"level":"info","msg":"Flags: region=us-east-1, interval=1200, profile=, secretName=ecr-docker-secret","time":"2018-09-29T16:48:17Z"}
{"level":"info","msg":"creating kubernetes secret","time":"2018-09-29T16:48:19Z"}
{"level":"info","msg":"updated kubernetes secret: ecr-docker-secret","time":"2018-09-29T16:48:19Z"}
kubectl get secrets ecr-docker-secret -n ${SECRET_NAMESPACE} -o json | jq '.data["config.json"]' | tr -d '"' | base64 --decode
{
"auths": {
"https://${AWS_PROFILE}.dkr.ecr.us-east-1.amazonaws.com": {
"auth": "....."
}
}
}
- Install Go
- Download the project
go get github.com/bzon/ecr-k8s-secret-creator
- Go to
$GOPATH/github.com/bzon/ecr-k8s-secret-creator
- Change the docker repository in the
Makefile
to your own docker repository - Run
make docker-build
- Run
make push