kubergrunt
is a standalone go binary with a collection of commands that attempts to fill in the gaps between Terraform,
Helm, and Kubectl for managing a Kubernetes Cluster.
Some of the features of kubergrunt
include:
- Configuring
kubectl
to authenticate with a given EKS cluster. Learn more about authenticatingkubectl
to EKS in the our production deployment guide. - Managing Helm and associated TLS certificates on any Kubernetes cluster.
- Setting up Helm client with TLS certificates on any Kubernetes cluster.
- Generating TLS certificate key pairs and storing them as Kubernetes
Secrets
on any Kubernetes cluster.
The binaries are all built as part of the CI pipeline on each release of the package, and is appended to the corresponding release in the Releases Page. You can download the corresponding binary for your platform from the releases page.
Alternatively, you can install kubergrunt
using the Gruntwork
Installer. For example, to install version v0.5.13
:
gruntwork-install --binary-name "kubergrunt" --repo "https://github.com/gruntwork-io/kubergrunt" --tag "v0.5.13"
Note that third-party Kubergrunt packages may not be updated with the latest version, but are often close. Please check your version against the latest available on the Releases Page.
Chocolatey (Windows):
choco install kubergrunt
You need to authenticate with the AWS CLI before you can run kubergrunt/eksctl commands, please see our authentication guide
The main package is in cmd
. To build the binary, you can run:
go build -o bin/kubergrunt ./cmd
If you need to set a version on the binary (so that kubergrunt --version
works), you use ldflags
to set the version
string on the compiled binary:
go build -o kubergrunt -ldflags "-X main.VERSION=v0.7.6 -extldflags '-static'" ./cmd
The following commands are available as part of kubergrunt
:
The eks
subcommand of kubergrunt
is used to setup the operator machine to interact with a Kubernetes cluster running
on EKS.
This subcommand verifies that the specified EKS cluster is up and ready. An EKS cluster is considered ready when:
- The cluster status reaches ACTIVE state.
- The cluster Kubernetes API server endpoint responds to http requests.
When passing --wait
to the command, this command will wait until the EKS cluster reaches the ready state, or it
times out. The timeout parameters are configurable with the --max-retries
and --sleep-between-retries
options, where
--max-retries
specifies the number of times the command will try to verify a specific condition before giving up, and
--sleep-between-retries
specifies the duration of time (e.g 10m = 10 minutes) to wait between each trial. So for
example, if you ran the command:
kubergrunt eks verify --eks-cluster-arn $EKS_CLUSTER_ARN --wait --max-retries 10 --sleep-between-retries 15s
and the cluster was not active yet, this command will query the AWS API up to 10 times, waiting 15 seconds inbetween each try for a total of 150 seconds (2.5 minutes) before timing out.
Run kubergrunt eks verify --help
to see all the available options.
Similar Commands:
- AWS CLI (
aws eks wait
): This command will wait until the EKS cluster reaches the ACTIVE state. Note that oftentimes the Kubernetes API endpoint has a delay in accepting traffic even after reaching the ACTIVE state. We have observed it take up to 1.5 minutes after the cluster becomes ACTIVE before we can have a valid TCP connection with the Kubernetes API endpoint.
This subcommand will setup the installed kubectl
with config contexts that will allow it to authenticate to a
specified EKS cluster by leveraging the kubergrunt eks token
command. This binary is designed to be used as part of
one of the modules in the package, although this binary supports running as a standalone binary. For example, this
binary might be used to setup a new operator machine to be able to talk to an existing EKS cluster.
For example to setup a kubectl
install on an operator machine to authenticate with EKS:
kubergrunt eks configure --eks-cluster-arn $EKS_CLUSTER_ARN
Run kubergrunt eks configure --help
to see all the available options.
Similar Commands:
- AWS CLI (
aws eks update-kubeconfig
): This command will configurekubeconfig
in a similar manner. Instead of usingkubergrunt eks token
, this version will use theget-token
subcommand built into the AWS CLI.
This subcommand is used by kubectl
to retrieve an authentication token using the AWS API authenticated with IAM
credentials. This token is then used to authenticate to the Kubernetes cluster. This command embeds the
aws-iam-authenticator
tool into kubergrunt
so that operators don't have to install a separate tool to manage
authentication into Kubernetes.
The configure
subcommand of kubergrunt eks
assumes you will be using this method to authenticate with the Kubernetes
cluster provided by EKS. If you wish to use aws-iam-authenticator
instead, replace the auth info clause of the kubectl
config context.
This subcommand also supports outputting the token in a format that is consumable by terraform as an external data
source when you pass in the --as-tf-data
CLI arg.
You can then pass the token directly into the kubernetes
provider configuration. For example:
# NOTE: Terraform does not allow you to interpolate resources in a provider config. We work around this by using the
# template_file data source as a means to compute the resource interpolations.
provider "kubernetes" {
load_config_file = false
host = "${data.template_file.kubernetes_cluster_endpoint.rendered}"
cluster_ca_certificate = "${base64decode(data.template_file.kubernetes_cluster_ca.rendered)}"
token = "${lookup(data.external.kubernetes_token.result, "token_data")}"
}
data "template_file" "kubernetes_cluster_endpoint" {
template = "${module.eks_cluster.eks_cluster_endpoint}"
}
data "template_file" "kubernetes_cluster_ca" {
template = "${module.eks_cluster.eks_cluster_certificate_authority}"
}
data "external" "kubernetes_token" {
program = ["kubergrunt", "--loglevel", "error", "eks", "token", "--as-tf-data", "--cluster-id", "${module.eks_cluster.eks_cluster_name}"]
}
This will configure the kubernetes
provider in Terraform without setting up kubeconfig, allowing you to do everything
in Terraform without side effects to your local machine.
Similar Commands:
- AWS CLI (
aws eks get-token
): This command will do the same thing, but does not provide any specific optimizations for terraform. - Terraform
aws_eks_cluster_auth
data source: This data source can be used to retrieve a temporary auth token for EKS in Terraform. This can only be used in Terraform. aws-iam-authenticator
: This is a standalone binary that can be used to fetch a temporary auth token.
This subcommand will take the EKS OIDC Issuer URL and retrieve the root CA thumbprint. This is used to set the trust relation for any certificates signed by that CA for the issuer domain. This is necessary to setup the OIDC provider, which is used for the IAM Roles for Service Accounts feature of EKS.
You can read more about the general procedure for retrieving the root CA thumbprint of an OIDC Provider in the official documentation.
To retrieve the thumbprint, call the command with the issuer URL:
kubergrunt eks oidc-thumbprint --issuer-url $ISSUER_URL
This will output the thumbprint to stdout in JSON format, with the key thumbprint
.
Run kubergrunt eks oidc-thumbprint --help
to see all the available options.
Similar Commands:
- You can use
openssl
to retrieve the thumbprint as described by the official documentation. eksctl
provides routines for directly configuring the OIDC provider so you don't need to retrieve the thumbprint.
This subcommand will initiate a rolling deployment of the current AMI config to the EC2 instances in your EKS cluster.
This command will not deploy or update an application deployed on your Kubernetes cluster (e.g Deployment
resource,
Pod
resource, etc). We provide helm charts that you can use to deploy your applications on to a Kubernetes cluster.
See our helm-kubernetes-services
repo for more info.
Instead, this command is for managing and deploying an update to the EC2 instances underlying your EKS cluster.
Terraform and AWS do not provide a way to automatically roll out a change to the Instances in an EKS Cluster. Due to Terraform limitations (see here for a discussion), there is currently no way to implement this purely in Terraform code. Therefore, we've created this subcommand that can do a zero-downtime roll out for you.
To deploy a change (such as rolling out a new AMI) to all EKS workers using this command:
- Make sure the
cluster_max_size
is at least twice the size ofcluster_min_size
. The extra capacity will be used to deploy the updated instances. - Update the Terraform code with your changes (e.g. update the
cluster_instance_ami
variable to a new AMI). - Run
terraform apply
. - Run the command:
kubergrunt eks deploy --region REGION --asg-name ASG_NAME
When you call the command, it will:
- Double the desired capacity of the Auto Scaling Group that powers the EKS Cluster. This will launch new EKS workers with the new launch configuration.
- Wait for the new nodes to be ready for Pod scheduling in Kubernetes. This includes waiting for the new nodes to be registered to any external load balancers managed by Kubernetes.
- Cordon the old instances in the ASG so that they won't schedule new Pods.
- Drain the pods scheduled on the old EKS workers (using the equivalent of
kubectl drain
), so that they will be rescheduled on the new EKS workers. - Wait for all the pods to migrate off of the old EKS workers.
- Set the desired capacity down to the original value and remove the old EKS workers from the ASG.
Note that to minimize service disruption from this command, your services should setup a PodDisruptionBudget, a readiness probe that fails on container shutdown events, and implement graceful handling of SIGTERM in the container. You can learn more about these features in our blog post series covering them.
Currently kubergrunt
does not implement any checks for these resources to be implemented. However in the future, we
plan to bake in checks into the deployment command to verify that all services have a disruption budget set, and warn
the user of any services that do not have a check.
eks deploy
recovery file
Due to the nature of rolling update, the deploy
subcommand performs multiple sequential actions that
depend on success of the previous operations. To mitigate intermittent failures, the deploy
subcommand creates a
recovery file in the working directory for storing current deploy state. The recovery file is updated after
each stage and if the deploy
subcommand fails for some reason, execution resumes from the last successful state.
The existing recovery file can also be ignored with the --ignore-recovery-file
flag. In this case the recovery
file will be re-initialized.
This subcommand will sync the core components of an EKS cluster to match the deployed Kubernetes version by following the steps listed in the official documentation.
The core components managed by this command are:
- kube-proxy
- Amazon VPC CNI plug-in
- CoreDNS
By default, this command will rotate the images without waiting for the Pods to be redeployed. You can use the --wait
option to force the command to wait until all the Pods have been replaced.
Example:
kubergrunt eks sync-core-components --eks-cluster-arn EKS_CLUSTER_ARN
This subcommand cleans up the leftover AWS-managed security groups that are associated with an EKS cluster you intend to destroy. It accepts
--eks-cluster-arn
: the ARN of the EKS cluster--security-group-id
: a known security group ID associated with the EKS cluster--vpc-id
: the VPC ID where the cluster is located
It also looks for other security groups associated with the EKS cluster, such as the security group created by the AWS Load Balancer Controller. To safely delete these resources, it detaches and deletes any associated AWS Elastic Network Interfaces.
Example:
kubergrunt eks cleanup-security-group --eks-cluster-arn EKS_CLUSTER_ARN --security-group-id SECURITY_GROUP_ID \
--vpc-id VPC_ID
This subcommand can be used to toggle the CoreDNS service between scheduling on Fargate and EC2 worker types. During
the creation of an EKS cluster that uses Fargate, schedule-coredns fargate
will annotate the deployment so that
CoreDNS can find and allow EKS to use Fargate nodes. To switch back to EC2, you can run schedule-coredns ec2
to
reset the annotations such that EC2 nodes can be found by CoreDNS.
This command is useful when creating Fargate only EKS clusters. By default, EKS will schedule the coredns
service
assuming EC2 workers. You can use this command to force the service to run on Fargate.
You can also use this command in local-exec
provisioners on an aws_eks_fargate_profile
resource so you can
schedule the CoreDNS service after creating the profile, and revert back when destroying the profile.
Currently fargate
and ec2
are the only subcommands that schedule-coredns
accepts.
Examples:
kubergrunt eks schedule-coredns fargate --eks-cluster-name EKS_CLUSTER_NAME --fargate-profile-arn FARGATE_PROFILE_ARN
kubergrunt eks schedule-coredns ec2 --eks-cluster-name EKS_CLUSTER_NAME --fargate-profile-arn FARGATE_PROFILE_ARN
This subcommand can be used to drain Pods from the instances in the provided Auto Scaling Groups. This can be used to gracefully retire existing Auto Scaling Groups by ensuring the Pods are evicted in a manner that respects disruption budgets.
You can read more about the drain operation in the official documentation.
To drain the Auto Scaling Group my-asg
in the region us-east-2
:
kubergrunt eks drain --asg-name my-asg --region us-east-2
You can drain multiple ASGs by providing the --name
option multiple times:
kubergrunt eks drain --asg-name my-asg-a --name my-asg-b --name my-asg-c --region us-east-2
The k8s
subcommand of kubergrunt
includes commands that directly interact with the Kubernetes resources.
This subcommand waits for the Ingress endpoint to be provisioned. This will monitor the Ingress resource, continuously checking until the endpoint is allocated to the Ingress resource or times out. By default, this will try for 5 minutes (max retries 60 and time betweeen sleep of 5 seconds).
You can configure the timeout settings using the --max-retries and --sleep-between-retries CLI args. This will check for --max-retries times, sleeping for --sleep-between-retries inbetween tries.
For example, if you ran the command:
kubergrunt k8s wait-for-ingress \
--ingress-name $INGRESS_NAME \
--namespace $NAMESPACE \
--max-retries 10 \
--sleep-between-retries 15s
this command will query the Kubernetes API to check the Ingress
resource up to 10 times, waiting for 15 seconds
inbetween each try for a total of 150 seconds (2.5 minutes) before timing out.
Run kubergrunt k8s wait-for-ingress --help
to see all the available options.
This subcommand will call out to kubectl with a temporary file that acts as the kubeconfig, set up with the parameters
--kubectl-server-endpoint
, --kubectl-certificate-authority
, --kubectl-token
. Unlike using kubectl directly, this
command allows you to pass in the base64 encoded certificate authority data directly as opposed to as a file.
To forward args to kubectl, pass all the args you wish to forward after a --
. For example, the following command runs
kubectl get pods -n kube-system
:
kubergrunt k8s kubectl \
--kubectl-server-endpoint $SERVER_ENDPOINT \
--kubectl-certificate-authority $SERVER_CA \
--kubectl-token $TOKEN \
-- get pods -n kube-system
Run kubergrunt k8s kubectl --help
to see all the available options.
The tls
subcommand of kubergrunt
is used to manage TLS certificate key pairs as Kubernetes Secrets.
This subcommand will generate new TLS certificate key pairs based on the provided configuration arguments. Once the certificates are generated, they will be stored on your targeted Kubernetes cluster as Secrets. This supports features such as:
- Generating a new CA key pair and storing the generated key pair in your Kubernetes cluster.
- Issuing a new signed TLS certificate key pair using an existing CA stored in your Kubernetes cluster.
- Replacing the stored certificate key pair in your Kubernetes cluster with a newly generated one.
- Controlling which Namespace the Secrets are stored in.
For example, to generate a new CA key pair, issue a TLS certificate key pair, storing each of those as the Secrets
ca-keypair
and tls-keypair
respectively:
# Generate the CA key pair
kubergrunt tls gen \
--namespace kube-system \
--secret-name ca-keypair \
--ca \
--tls-common-name kiam-ca \
--tls-org Gruntwork \
--tls-org-unit IT \
--tls-city Phoenix \
--tls-state AZ \
--tls-country US \
--secret-annotation "gruntwork.io/version=v1"
# Generate a signed TLS key pair using the previously created CA
kubergrunt tls gen \
--namespace kube-system \
--secret-name tls-keypair \
--ca-secret-name ca-keypair \
--tls-common-name kiam-server \
--tls-org Gruntwork \
--tls-org-unit IT \
--tls-city Phoenix \
--tls-state AZ \
--tls-country US \
--secret-annotation "gruntwork.io/version=v1"
The first command will generate a CA key pair and store it as the Secret ca-keypair
. The --ca
argument signals to
kubergrunt
that the TLS certificate is for a CA.
The second command uses the generated CA key pair to issue a new TLS key pair. The --ca-secret-name
signals
kubergrunt
to use the CA key pair stored in the Kubernetes Secret ca-keypair
.
This command should be run by a cluster administrator to ensure access to the Secrets are tightly controlled.
See the command help for all the available options: kubergrunt tls gen --help
.
The helm
subcommand contained utilities for managing Helm v2, and is not necessary for Helm v3. This subcommand was
removed as of kubergrunt
version v0.6.0
with Helm v2 reaching end of life.
kubergrunt
is maintained by Gruntwork. If you are looking for help or commercial support,
send an email to [email protected].
Gruntwork can help with:
- Setup, customization, and support for this project.
- Modules and submodules for other types of infrastructure in major cloud providers, such as VPCs, Docker clusters, databases, and continuous integration.
- Modules and Submodules that meet compliance requirements, such as HIPAA.
- Consulting & Training on AWS, GCP, Terraform, and DevOps.
Contributions are very welcome! Check out the Contribution Guidelines for instructions.
This project follows the principles of Semantic Versioning. You can find each new release, along with the changelog, in the Releases Page.
During initial development, the major version will be 0 (e.g., 0.x.y
), which indicates the code does not yet have a
stable API. Once we hit 1.0.0
, we will make every effort to maintain a backwards compatible API and use the MAJOR,
MINOR, and PATCH versions on each release to indicate any incompatibilities.
Please see LICENSE and NOTICE for how the code in this repo is licensed.
Copyright © 2020 Gruntwork, Inc.