Consul and Kubernetes deployment guide
This tutorial covers the necessary steps to install and configure a new Consul datacenter on Kubernetes, as defined in the Consul Reference Architecture tutorial. By the end of this tutorial, you will be able to:
- Identify installation prerequisites
- Customize the Helm chart to fit your environmental requirements
- Interact with your new Consul datacenter
Prerequisites
To complete this tutorial you will need:
- Access to a Kubernetes cluster
- kubectl
- helm v3.2.1+ or Consul-K8S-CLI
Configure Kubernetes permissions
Before deploying Consul, you will need to have kubectl
credentials that allow
you to create and modify policies, deploy services, create secrets, and create
RBAC objects. You can find documentation for RBAC and service accounts for the
following cloud providers using the links below.
Note
Consul can be deployed on any properly configured Kubernetes cluster in the cloud or on-premises.
Configure Kubernetes secrets
You may wish to create Kubernetes secrets to store Consul data. Secrets can be referenced in the customized Helm chart values file.
Add an Enterprise license key (optional)
If you have purchased Enterprise Consul, you will need to add the enterprise license
to the Kubernetes secretes engine. Also, you will need to specify Consul image
using the Helm chart values file. The default image used by the official Consul
helm chart is the open source version of Consul. The image name format will resemble
this string: hashicorp/consul-enterprise:<latest version>-ent
.
For Enterprise customers that have purchased a license, work with your Customer Success Manager (CSM) to get the appropriate license. Once you have obtained the license, you can use the following command to register it as a secret in Kubernetes.
Configure Helm chart
Now that you have prepared your Kubernetes cluster, you can customize the Helm chart. First, you will need to add the HashiCorp Helm Chart repository:
The Helm chart is configured with default values. You can override these values
by creating your own config.yaml
file and passing it to helm install
with the
-f
flag, e.g. helm install consul hashicorp/consul -f config.yaml
. Below
we detail some parameters you can customize, and provide an example
file. However, you should consider your particular production needs when
configuring your chart.
Global key
The values under the global key will affect all the other parameters in the chart.
To enable all Consul components in the Helm chart, set enabled
to
true
. This means servers, clients, Consul DNS, and the Consul UI will be
installed with their defaults. You may also set the following global
parameters based on your specific environment requirements or naming conventions.
image
is the name and tag of the Consul Docker image - override this if you want to run a specific version of Consul, or if you are using Consul Enterprise. Consult the Consul compatibility matrix before overriding the image setting to make sure the chart version you are using is compatible with the image you are specifying.datacenter
the name of your Consul datacenter. The default value isdc1
, but can be overridden if necessary.domain
the domain Consul uses for DNS queries. The default value isconsul
but can be overridden if necessary.
For security, we recommend you set the global.acls.manageSystemACLs
parameter
to true. This will instruct Consul to initialize it's ACL system during chart installation.
Additionally, setting the global.gossipEncryption.autoGenerate
parameter to true will configure gossip encryption automatically with a random key when Consul is installed.
We also recommend you read the Consul Helm chart documentation to review all the possible global parameters.
Consul UI
To enable the Consul web UI update the ui
section to your values file and set
enabled
to true
.
Note
You can also set up a LoadBalancer resource or other service type in Kubernetes to make it easier to access the UI.
Consul servers
For production deployments, you should deploy 3 or 5 Consul servers for quorum and failure tolerance. For most deployments, 3 servers are adequate.
In the server section set replicas
to 3. This will deploy three servers and
cause Consul to wait to perform leader election until all three are healthy. The
resources
will depend on your environment; in the example at the end of the
tutorial, the resources are set for a large environment.
Affinity
To ensure the Consul servers are placed on different Kubernetes nodes, you will
need to configure affinity. Otherwise, the failure of one Kubernetes node could
cause the loss of multiple Consul servers, and result in quorum loss. The
default values.yaml
has affinity configured correctly.
Enterprise license
If you have an Enterprise license,
you should reference the Kubernetes secret in the enterpriseLicense
parameter.
Read the Consul Helm chart documentation to review the server parameters related to configuring an enterprise license.
Consul clients
Since a Consul client is deployed on every Kubernetes node, you do not need to specify the number of clients for your deployments. You will, however, need to specify resources. Since Consul clients are designed for horizontal scalability, the resources in the example at the end of this tutorial should be sufficient for most production scenarios.
It is worth mentioning that when configuring clients, you have the ability to
override the default node reconnect timeout. By default, Consul will wait 72 hours
before reaping unresponsive nodes. Since Consul on Kubernetes often runs on ephemeral
infrastructure, such as spot instances, this default setting may not meet your
needs. To override the default, you can set the client.extraConfig
setting by
passing a valid JSON object that includes an advertise_reconnect_timeout
key
set to the timeout interval that makes sense for your environment. As shown in
this example.
To review all the possible client parameters, refer to the Consul Helm chart documentation.
Consul service mesh configuration
To enable Consul service mesh, set enabled
under the connectInject
key to
true
. In the connectInject
section, you can also configure security
features. Enabling the default
parameter will result in the injector automatically
injecting the envoy-sidecar
sidecar proxy and consul-sidecar
life-cycle sidecar
into all pods. If you prefer to manually annotate which pods to inject, you should
set connectInject.default
to false
.
Setting the aclBindingRuleSelector
parameter to serviceaccount.name!=default
ensures that new services do not all receive the same token if you are only using
a default service account. This setting is only necessary if you have enabled ACLs
in the global section.
Read more about the Connect Inject parameters.
Complete example
Your finished config.yaml
file should resemble the following example. For more
complete descriptions of all the available parameters, check the values.yaml
file provided with the Helm chart and the reference documentation.
Deploy Consul
Now that you have customized your config.yaml
file, you can deploy Consul with Helm or the Consul K8S CLI. It is recommended to deploy Consul into its own dedicated namespace as shown below. This should only take a few minutes. The Consul pods should appear in the pod list immediately.
To check the deployment process on the command line you can use kubectl get pods --namespace consul
.
Next steps
In this tutorial, you configured Consul using the Helm chart for a production environment. This involved ensuring that your Kubernetes cluster had a properly distributed configuration for Consul servers, specifying enough resources for your agents, securing the datacenter with ACLs and gossip encryption, and enabling other Consul functionality including Connect service mesh and the Consul UI.
Now, you can interact with your Consul datacenter through the UI or CLI.
If you exposed the UI using a load balancer it will be available at the
LoadBalancer Ingress
IP address and Port
that is output from the following
command.
To access the Consul CLI, open a terminal session using the Kubernetes CLI.
Note
You may need to replace consul-server-0 with the server name from your datacenter if you overrode the global.name
setting in your Helm values file.
This will allow you to navigate the file system and run Consul CLI commands on the pod. For example you can view the Consul members.
When you have finished interacting with the pod, exit the shell.
Next steps
In this tutorial, you installed and configured a new Consul datacenter on Kubernetes using the official Helm chart or Consul K8S CLI. To continue learning about Consul service discovery and service mesh, complete the Consul service mesh tutorials. For more information on securing and debugging Consul on Kubernetes review the Secure Consul and Registered Services on Kubernetes tutorial.