Skip to content

Latest commit

 

History

History
359 lines (249 loc) · 12.8 KB

README.md

File metadata and controls

359 lines (249 loc) · 12.8 KB

kubernetes-conjur-deploy

This repository contains scripts for automating the deployment of Conjur followers to a Kubernetes or OpenShift environment. These scripts can also be used to deploy a full cluster with Master and Standbys for testing and demo purposes but this is not recommended for a production deployment of Conjur.

Enterprise Only. To deploy Conjur OSS, please use the Conjur OSS helm chart.


Setup

The Conjur deployment scripts pick up configuration details from local environment variables. The setup instructions below walk you through the necessary steps for configuring your environment and show you which variables need to be set before deploying.

All environment variables can be set/defined with:

  • bootstrap.env file if deploying the follower to Kubernetes or OpenShift
  • dev-bootstrap.env for all other configurations.

Edit the values per instructions below, source the appropriate file and run 0_check_dep0endencies.sh to verify.

The Conjur appliance image can be loaded with _load_conjur_tarfile.sh. The script uses environment variables to locate the tarfile image and the value to use as a tag once it's loaded.

Conjur Configuration

Appliance Image

You need to obtain a Docker image of the Conjur appliance and push it to an accessible Docker registry. Provide the image and tag like so:

export CONJUR_APPLIANCE_IMAGE=<tagged-docker-appliance-image>

You will also need to provide an ID for the Conjur authenticator that will later be used in Conjur policy to provide your apps with access to secrets through Conjur:

export AUTHENTICATOR_ID=<authenticator-id>

This ID should describe the cluster in which Conjur resides. For example, if you're hosting your dev environment on GKE you might use gke/dev.

Follower Seed

You will need to provide a follower seed file generated from your Conjur Master. The seed can be generated by SSH-ing into your Master and running:

NOTE: If you are running this code to deploy a follower that will run in a separate cluster from the master, you must force-generate the follower certificate manually before creating the seed to prevent the resulting certificate from omitting the future in-cluster subject altname included.

To generate a follower seed with the appropriate in-cluster subject altname for followers that are not in the same cluster as master, we first will need to issue a certificate on master. Skip this step if master is collocated with the follower.

$ evoke ca issue --force <follower_external_fqdn> conjur-follower.<conjur_namespace_name>.svc.cluster.local

We now need to create the seed archive with the proper information:

$ evoke seed follower <follower_external_fqdn> > /tmp/follower-seed.tar

If you are on the same node as the master container, you can also export the seed with:

$ sudo docker exec <container_id> evoke seed follower <follower_external_fqdn> > /tmp/follower-seed.tar

Note: the exported seed file will not be copied to host properly if you use -t flag with the docker exec command.

Copy the resulting seed file from the Conjur master to your local filesystem and set the following environment variable to point to it:

export FOLLOWER_SEED_PATH=path/to/follower/seed/file

The deploy scripts will copy the seed to your follower pods and use it to configure them as Conjur followers.

Important note: Make sure to delete any copies of the seed after use. It contains sensitive information and can always be regenerated on the Master.

Platform Configuration

If you are working with OpenShift, you will need to set:

export PLATFORM=openshift
export OSHIFT_CLUSTER_ADMIN_USERNAME=<name-of-cluster-admin> # system:admin in minishift
export OSHIFT_CONJUR_ADMIN_USERNAME=<name-of-conjur-namespace-admin> # developer in minishift

Otherwise, $PLATFORM variable will default to kubernetes.

Before deploying Conjur, you must first make sure that you are connected to your chosen platform with a user that has the cluster-admin role. The user must be able to create namespaces and cluster roles.

Conjur Namespace

Provide the name of a namespace in which to deploy Conjur:

export CONJUR_NAMESPACE_NAME=<my-namespace>

The conjur-authenticator Cluster Role

Conjur's Kubernetes authenticator requires the following privileges:

  • ["get", "list"] on "pods" for confirming a pod's namespace membership
  • ["create", "get"] on "pods/exec" for injecting a certificate into a pod

The deploy scripts include a manifest that defines the conjur-authenticator cluster role, which grants these privileges. Create the role now (note that your user will need to have the cluster-admin role to do so):

# Kubernetes
kubectl apply -f ./kubernetes/conjur-authenticator-role.yaml

# OpenShift
oc apply -f ./openshift/conjur-authenticator-role.yaml

Docker Configuration

Install Docker on your local machine if you do not already have it.

Kubernetes

You will need to provide the domain and any additional pathing for the Docker registry from which your Kubernetes cluster pulls images:

export DOCKER_REGISTRY_URL=<registry-domain>
export DOCKER_REGISTRY_PATH=<registry-domain>/<additional-pathing>

Note that the deploy scripts will be pushing images to this registry so you will need to have push access.

If you are using a private registry, you will also need to provide login credentials that are used by the deployment scripts to create a secret for pulling images:

export DOCKER_USERNAME=<your-username>
export DOCKER_PASSWORD=<your-password>
export DOCKER_EMAIL=<your-email>

Please make sure that you are logged in to the registry before deploying.

OpenShift

OpenShift users should make sure the integrated Docker registry in your OpenShift environment is available and that you've added it as an insecure registry in your local Docker engine. You must then specify the path to the OpenShift registry like so:

export DOCKER_REGISTRY_PATH=docker-registry-<registry-namespace>.<routing-domain>

Please make sure that you are logged in to the registry before deploying.

Running OpenShift in Minishift

You can use Minishift to run OpenShift locally in a single-node cluster. Minishift provides a convenient way to test out Conjur deployments on a laptop or local machine and also provides an integrated Docker daemon from which to stage and push images into the OpenShift registry. The ./openshift subdirectory contains two files:

  • _minishift-boot.env that defines environment variables to configure Minishift, and
  • _minishift-start.sh to startup Minishift. The script assumes VirtualBox as the hypervisor but others are supported. See https://github.com/minishift/minishift for more information.

Steps to startup Minishift:

  1. Ensure VirtualBox is installed
  2. cd openshift
  3. Run ./minishift-start.sh
  4. source minishift.env to gain use of the internal Docker daemon
  5. cd ..
  6. Use dev-bootstrap.env for your variable configuration
  7. Run ./start

Usage

Deploying Conjur Follower

Ensure that bootstrap.env has the FOLLOWER_SEED_PATH variable set to the seed file created here.

After verifying this setting, source ./bootstrap.env and then run ./start to execute the scripts necessary to have the follower deployed in your environment.


Deploying Conjur Master and Followers (Test and Demo Only)

Master Cluster configuration

Please note that running master cluster in OpenShift and Kubernetes environments is not recommended and should be only done for test and demo setups.

As mentioned before if you are using these scripts to deploy a full cluster, you will need to set in dev-bootstrap.env:

export DEPLOY_MASTER_CLUSTER=true

You will also need to set a few environment variable that are only used when configuring the Conjur master. If you are working with Conjur v4, you will need to set:

export CONJUR_VERSION=4

along with any other changes you might want.

Otherwise, this variable will default to 5.

You must also provide an account name and password for the Conjur admin account:

export CONJUR_ACCOUNT=<my_account_name>
export CONJUR_ADMIN_PASSWORD=<my_admin_password>

Finally, run ./start to execute the scripts necessary for deploying Conjur.

Data persistence

The Conjur master and standbys are deployed as a Stateful Set on supported target platforms (Kubernetes 1.5+ / OpenShift 3.5+). Database and configuration data is symlinked and mounted to persistent volumes. These manifests assume a default Storage Class is set up for the cluster so persistent volume claims will be fulfilled.

Volumes:

  • /opt/conjur/dbdata - 2GB, database persistence
  • /opt/conjur/data - 1GB, seed file persistence

Setup

To configure the Conjur master to persist data, run these commands in the Conjur master container before running evoke configure master ....

# mv /var/lib/postgresql/9.3 /opt/conjur/dbdata/
# ln -sf /opt/conjur/dbdata/9.3 /var/lib/postgresql/9.3

# evoke seed standby > /opt/conjur/data/standby-seed.tar

Note that setup is done as part of script 4_configure_master.sh.

Restore

If the Conjur master pod is rescheduled the persistent volumes will be reattached. Once the pod is running again, run these commands to restore the master.

# rm -rf /var/lib/postgresql/9.3
# ln -sf /opt/conjur/dbdata/9.3 /var/lib/postgresql/9.3

# cp /opt/conjur/data/standby-seed.tar /opt/conjur/data/standby-seed.tar-bkup
# evoke unpack seed /opt/conjur/data/standby-seed.tar
# cp /opt/conjur/data/standby-seed.tar-bkup /opt/conjur/data/standby-seed.tar
# rm /etc/chef/solo.json

# evoke configure master ...  # using the same arguments as the first launch

Standbys must also be reconfigured since the Conjur master pod IP changes.

Run relaunch_master.sh to try this out in your cluster, after running the deploy. Our plan is to automate this process with a Kubernetes operator.

Conjur CLI

The deploy scripts include a manifest for creating a Conjur CLI container within the Kubernetes environment that can then be used to interact with Conjur. Deploy the CLI pod and SSH into it:

# Kubernetes
kubectl create -f ./kubernetes/conjur-cli.yaml
kubectl exec -it [cli-pod-name] bash

# OpenShift
oc create -f ./openshift/conjur-cli.yaml
oc exec -it <cli-pod-name> bash

Once inside the CLI container, use the admin credentials to connect to Conjur:

conjur init -h conjur-master

Follow our CLI usage instructions to get started with the Conjur CLI.

Conjur UI

Visit the Conjur UI URL in your browser and login with the admin credentials to access the Conjur UI.


Test App Demo

The kubernetes-conjur-demo repo deploys test applications that retrieve secrets from Conjur and serves as a useful reference when setting up your own applications to integrate with Conjur.

Auto Enrolled Follower

  1. Create a follower seed file on the Conjur master. Run:

    evoke seed follower <conjur-follower-service-name> <master load balancer public DNS> > follower-seed.tar

  2. Copy the seed file to your local machine.

  3. Create a secret from the follower seed file:

    oc create secret generic conjur-follower-seed-secret --from-file=/path/to/follower-seed.tar --namespace=<conjur namespace>

  4. Modify the conjur-follower-auto-enrolled.yaml with your:

  • Seed file secret name
  • Follower deployment config name
  • Follower service name
  1. Create an image pull secret from the the Conjur service account token:
  • Get the Conjur service account token name: sa_token=$(oc describe sa conjur-cluster | grep -i tokens | awk '{print $2}')
  • Get the token secret: sa_token_secret=$(oc describe secret $sa_token | grep -i token: | awk '{print $2}')
  • Save the secret as the deployment image pull secret:
    oc create secret docker-registry dockerpullsecret \
    --docker-server=${DOCKER_REGISTRY_PATH} \
    --docker-username=_ \
    --docker-password=$sa_token_secret \
    --docker-email=_

    oc secrets link serviceaccount/conjur-cluster secrets/dockerpullsecret --for=pull
  1. Create the Conjur follower deployment by running: oc create -f openshift/conjur-follower-auto-enrolled.yaml