Hardening The Deployment

To harden the demo deployment there are a number of steps to follow. The documentation below requires modifying the configuration of the Helm charts installed by following the Manual Install steps or running the deploy.sh script. To obtain the installed configuration for each installed Helm chart, run the following commands:

helm -n nshieldkeysafe5 get values --all --output yaml keysafe5-backend > keysafe5-backend-values.yaml
helm -n nshieldkeysafe5 get values --all --output yaml keysafe5-ui > keysafe5-ui-values.yaml
helm -n nshieldkeysafe5 get values --all --output yaml keysafe5-istio > keysafe5-istio-values.yaml

Documentation for each configurable value in the KeySafe 5 Helm charts can be found by untarring the chart.tgz and viewing the contents of either README.md or the default values.yaml file.

Certificates

The Manual Install steps and deploy.sh script will generate a local Certificate Authority, and install a number of short-lived demo certificates. These certificates must be replaced to continue using the system after their expiry.

Your new certificates will need to adhere to X.509 v3, sometimes known as a multiple-domain certificates, or SAN certificates. The X.509 extension Subject Alternative Name (SAN) allows specifying multiple hostnames, and has replaced Common Name as the source of the hostname.

External KeySafe 5 Server TLS Certificate

To update the TLS certificate used for the KeySafe 5 Server (for HTTPS connections to the REST API or User Interface) you must create a Kubernetes Secret containing the new certificate/private key and redeploy the keysafe5-istio Helm chart.

For more information on enabling HTTPS for helm-keysafe5-istio, see the Helm Chart Installation section of the Installation Guide.

The Manual Install steps and deploy.sh script create a Kubernetes Secret called keysafe5-server-credential. You can either delete the existing Secret as shown below, or use a different name for the Secret containing your new TLS certificate.

kubectl --namespace istio-system delete secret keysafe5-server-credential

kubectl --namespace istio-system create secret tls keysafe5-server-credential \
  --cert=path/to/cert/file \
  --key=path/to/key/file

Before running helm upgrade, set the following values in your keysafe5-istio-values.yaml:

httpsEnabled=true
tls.existingSecret=keysafe5-server-credential (or the name you used when creating the Kubernetes Secret containing your certificate/private key)

helm upgrade --install keysafe5-istio \
  --namespace=nshieldkeysafe5 \
  --values keysafe5-istio-values.yaml \
  --wait --timeout 1m \
helm-charts/nshield-keysafe5-istio-1.3.0.tgz

Internal Certificates

The Manual Install steps and deploy.sh script create a Certificate Authority, and install KeySafe 5 using TLS authorised by the CA for the communications between the central platform and MonogDB/RabbitMQ.

You can refresh these internal certificates by running the updateinternalcerts.sh script, specifying the number of days for which the new certificates will be valid.

The help for updateinternalcerts.sh is shown with the -h option:

  Usage: ./updateinternalcerts.sh [OPTION]... [SERVER] DAYS
  Generate certificates for SERVER that expire in DAYS days and update the
  KeySafe 5 internal TLS certificates for a deployment created by deploy.sh

  SERVER                        One of "mongodb" or "rabbitmq".
                                If the server is not specified then both will
                                be updated.
  DAYS                          Number of days before the generated certificates
                                expire
  -n, --namespace=NAMESPACE     The deploy script normally uses nshieldkeysafe5
                                as the namespace for the KeySafe 5 servers,
                                mongons for MongoDB, and rabbitns for RabbitMQ.
                                However you can override these to use a single
                                namespace for all the servers.  If that is the
                                case, then use this option.

  Certificate Addresses

  The certificates contain names and IP addresses that verify the connection.

  The backend servers use Kubernetes internal addressing to access the servers,
  and this script will set these up.

  MongoDB is only used by the backend services so does not require an external
  address.

  RabbitMQ is also accessed by the KeySafe 5 Agents through an external address.
  This script will use the IP address provided by the loadbalancer, but you may
  override this with a different IP address or DNS name.

  -r, --rabbitmq=ADDRESS        The external address for RabbitMQ
  -m, --mongodb=ADDRESS         The external address for MongoDB

  When you update the certificates for RabbitMQ the script will generate an
  updated agent-config.tar.gz and agent-config.zip files containing the updated
  TLS certificates. You must then update all nShield client machines with a
  KeySafe 5 agent installed to use the updated certificates and restart the
  agent so that the new configuration is applied.

  Examples

  Refresh certificates for both servers for the next 30 days, and set a DNS
  entry for RabbitMQ
    ./updateinternalcerts.sh -r rabbitmq.example.com 30

  Refresh just MongoDB for the next 4 weeks
    ./updateinternalcerts.sh mongodb 28

This script will update both MongoDB and RabbitMQ TLS certificates, though you may choose to update only one set of TLS certificates by specifying only that server.

The updateinternalcerts.sh script must be run from a directory containing the KeySafe 5 Helm charts (for example, from the root directory of the untarred KeySafe 5 package). If the CA, created when the original deploy script was run, is not available then a new one will be created and used.

If both certificates have expired when running updateinternalcerts.sh, updating the certificates of only one service may return an error. In this case re-running updateinternalcerts.sh without specifying any server will update both server certificates at once, and will usually solve the problem.

If an error occurs during certificate update you can restore the previous setup by rolling back Helm chart installations to a previous release, see Helm Chart Upgrade in the Upgrade section of the Installation Guide.

MongoDB TLS Certificates

The Manual Install steps and deploy.sh script installs the Bitnami MongoDB Helm chart with TLS enabled for the connections between KeySafe 5 and the MongoDB server and also with TLS used for authentication. These certificates will initially be valid for 30 days from the time the process was run.

You can refresh the MongoDB certificates by running the updateinternalcerts.sh script, specifying the number of days the new certificates will be valid for.

updateinternalcerts.sh mongodb 365

This script will:

Generate the new TLS certificates
Update the MongoDB helm chart to use the new certificates
Update the keysafe5-backend helm chart to use the new certificates

You may specify an external address with the -m parameter but this is usually not required.

RabbitMQ

The Manual Install steps and deploy.sh script installs the Bitnami RabbitMQ Helm chart with TLS enabled.

You can refresh the RabbitMQ certificates by running the updateinternalcerts.sh script, specifying the number of days the new certificates will be valid for.

updateinternalcerts.sh rabbitmq 365

This script will:

Generate the new TLS certificates
Update the RabbitMQ helm chart to use the new certificates
Update the keysafe5-backend helm chart to use the new certificates
Output a new agent-config.tar.gz that contains the agent configuration file and TLS certificates for authentication to RabbitMQ

The external address that will be put into the certificate, and the configuration file, will be the IP address of the Ingress service provided by the Kubernetes server. If you wish to use a different IP address or DNS name, this can be passed in with the -r parameter.

You will need to update your client machines with a KeySafe 5 agent installed to use the updated config and restart the agent so that the new configuration is applied.

sudo tar xf agent-config.tar.gz -C /opt/nfast/keysafe5/
/opt/nfast/scripts/init.d/keysafe5-agent restart

Authentication

If you chose to install the demo deployment without authentication you should enable authentication for accessing the KeySafe 5 REST API and User Interface.

For how to configure authentication for the KeySafe 5 REST APIs see Helm Chart Installation: helm-keysafe5-istio authentication in the Installation Guide.

To update the keysafe5-istio Helm chart installed by the demo deployment, set the following values in keysafe5-istio-values.yaml.

requireAuthn=true
issuer[0].authIssuer="https://foobar.auth0.com"
issuer[0].authJWKsURI="https://www.googleapis.com/oauth2/v1/certs"
issuer[0].authAudiences[0]="https://keysafe5.location"

Then run helm upgrade.

helm upgrade --install keysafe5-istio \
  --namespace=nshieldkeysafe5 \
  --values keysafe5-istio-values.yaml \
  --wait --timeout 1m \
helm-charts/nshield-keysafe5-istio-1.3.0.tgz

To update the keysafe5-ui Helm chart installed by the demo deployment, set the following values in keysafe5-ui-values.yaml:

authMethod=oidc

Untar the chart and copy your OIDC provider config file (OIDCProviders.json) into the config directory:

For more details on how to populate OIDCProviders.json and how to configure authentication for the KeySafe 5 User Interface see Helm Chart Installation: Configure UI authentication in the Installation Guide.

tar -xf helm-charts/nshield-keysafe5-ui-1.3.0.tgz -C helm-charts
cp my-oidc-provider-config.json helm-charts/nshield-keysafe5-ui/config/OIDCProviders.json

Then run helm upgrade:

helm upgrade --install keysafe5-ui \
  --namespace=nshieldkeysafe5 \
  --values keysafe5-ui-values.yaml \
  --wait --timeout 3m \
helm-charts/nshield-keysafe5-ui

K3s

If not using your own Kubernetes cluster, the deploy.sh script will create one using K3s. To harden this K3s install follow the official documentation at K3s Hardening Guide.

The deploy.sh script installs K3s with traefik and metrics-server explicitly disabled.