»Syncing Kubernetes and Consul Services
The services in Kubernetes and Consul can be automatically synced so that Kubernetes services are available to Consul agents and services in Consul can be available as first-class Kubernetes services. This functionality is provided by the consul-k8s project and can be automatically installed and configured using the Consul Helm chart.
Why sync Kubernetes services to Consul? Kubernetes services synced to the Consul catalog enable Kubernetes services to be accessed by any node that is part of the Consul cluster, including other distinct Kubernetes clusters. For non-Kubernetes nodes, they can access services using the standard Consul DNS or HTTP API.
Why sync Consul services to Kubernetes? Syncing Consul services to Kubernetes services enables non-Kubernetes services (such as external to the cluster) to be accessed in a native Kubernetes way: using kube-dns, environment variables, etc. This makes it very easy to automate external service discovery, including hosted services like databases.
»Installation and Configuration
The service sync is done using an external long-running process in the consul-k8s project. This process can run either in or out of a Kubernetes cluster. However, running this within the Kubernetes cluster is generally easier since it is automated using the Helm chart.
The Consul server cluster can run either in or out of a Kubernetes cluster. The Consul server cluster does not need to be running on the same machine or same platform as the sync process. The sync process needs to be configured with the address to a Consul agent as well as any additional access information such as ACL tokens.
To install the sync process, enable the catalog sync feature using
Helm values and
upgrade the installation using
helm upgrade for existing installs or
helm install for a fresh install.
syncCatalog: enabled: true
This will enable services to sync in both directions. You can also choose to only sync Kubernetes services to Consul or vice versa by disabling a direction.
To only enable syncing Consul services to Kubernetes use the config:
syncCatalog: enabled: true toConsul: false toK8S: true
To only enable syncing Kubernetes services to Consul use:
syncCatalog: enabled: true toConsul: true toK8S: false
See the Helm configuration for more information.
The sync process must authenticate to both Kubernetes and Consul to read and write services.
consul-k8s using the Helm chart then this authentication is handled for you.
consul-k8s outside of Kubernetes, a valid kubeconfig file must be provided with cluster
and authentication information. The sync process will look into the default locations
for both in-cluster and out-of-cluster authentication. If
then the sync program should work.
For Consul, if ACLs are configured on the cluster, a Consul
will need to be provided. Review the ACL rules
when creating this token so that it only allows the necessary privileges. The catalog
sync process accepts this token by using the
environment variable. This token should be set as a
and referenced in the Helm chart.
»Kubernetes to Consul
This sync registers Kubernetes services to the Consul catalog automatically.
This enables discovery and connection to Kubernetes services using native Consul service discovery such as DNS or HTTP. This is particularly useful for non-Kubernetes nodes. This also causes all discoverable services to be part of a central service catalog in Consul for further syncing into alternate Kubernetes clusters or other platforms.
»Kubernetes Service Types
Not all Kubernetes services are externally accessible. The sync program by default will only sync services of the following types or configurations. If a service type is not listed below, then the sync program will ignore that service type.
NodePort services register a static port that every node in the K8S cluster listens on.
For NodePort services, a Consul service instance will be created for each node that has the representative pod running. While Kubernetes configures a static port on all nodes in the cluster, this limits the number of service instances to be equal to the nodes running the target pods.
The service instances will be registered to the Kubernetes node name that each instance lives on. This is guaranteed unique by Kubernetes. An existing node entry will be used if it is already part of the Consul cluster (for example if you're running a client agent on all Kubernetes nodes). This allows the normal agent health checks for that node to continue working.
For LoadBalancer services, a single service instance will be registered with the external IP of the created load balancer. Because this is already a load balancer, only one service instance will be registered with Consul rather than registering each individual pod endpoint.
Any service type may specify an "external IP" configuration. The external IP must be configured by some other system, but any service discovery will resolve to this set of IP addresses rather than a virtual IP.
If an external IP list is present, a service instance in Consul will be created for each external IP. It is assumed that if an external IP is present that it is routable and configured by some other system.
ClusterIP services are synced by default as of
consul-k8s version 0.3.0. In
many Kubernetes clusters, ClusterIPs may not be accessible outside of the cluster,
so you may end up with services registered in Consul that are not routable. To
skip syncing ClusterIP services, set
false in the Helm chart values file.
By default, all valid service types (as explained above) are synced from every Kubernetes
namespace (except for
If you wish to only sync specific services via annotation, set the default to
syncCatalog: enabled: true default: false
And explicitly enable syncing specific services via the
kind: Service apiVersion: v1 metadata: name: my-service annotations: 'consul.hashicorp.com/service-sync': 'true'
NOTE: If the annotation is set to
false when the default sync is
true, the service will not be synced.
You can allow or deny syncing from specific Kubernetes namespaces by setting the
syncCatalog: enabled: true default: true k8sAllowNamespaces: ['*'] k8sDenyNamespaces: ['kube-system', 'kube-public']
In the default configuration (shown above), services from all namespaces except for
kube-public will be synced.
If you wish to only sync from specific namespaces, you can list only those
namespaces in the
syncCatalog: enabled: true default: true k8sAllowNamespaces: ['my-ns-1', 'my-ns-2'] k8sDenyNamespaces: 
If you wish to sync from every namespace except specific namespaces, you can
* in the allow list and then specify the non-syncing namespaces in the deny list:
syncCatalog: enabled: true default: true k8sAllowNamespaces: ['*'] k8sDenyNamespaces: ['no-sync-ns-1', 'no-sync-ns-2']
NOTE: The deny list takes precedence over the allow list. If a namespace is listed in both lists, it will not be synced.
When a Kubernetes service is synced to Consul, the name of the service in Consul by default will be the value of the "name" metadata on that Kubernetes service. This makes it so that service sync works with zero configuration changes. This can be overridden using an annotation to specify the Consul service name:
kind: Service apiVersion: v1 metadata: name: my-service annotations: 'consul.hashicorp.com/service-name': my-consul-service
If a conflicting service name exists in Consul, the sync program will register additional instances to that same service. Therefore, services inside and outside of Kubernetes should have different names unless you want either side to potentially connect. This default behavior also enables gracefully transitioning a service from outside of K8S to inside, and vice versa.
When syncing the Kubernetes service to Consul, the Consul service port will be the first defined port in the service. Additionally, all ports will be registered in the service instance metadata with the key "port-X" where X is the name of the port and the value is the externally accessible port.
The default service port can be overridden using an annotation:
kind: Service apiVersion: v1 metadata: name: my-service annotations: 'consul.hashicorp.com/service-port': 'http'
The annotation value may be a name of a port (recommended) or an exact port value.
A service registered in Consul from Kubernetes will always have the tag "k8s" added to it. Additional tags can be specified with a comma-separated annotation value as shown below. This will also automatically include the "k8s" tag which can't be disabled. The values should be specified comma-separated without any additional whitespace.
kind: Service apiVersion: v1 metadata: name: my-service annotations: 'consul.hashicorp.com/service-tags': 'primary,foo'
A service registered in Consul from Kubernetes will set the
external-source key to
"kubernetes". This can be used by API consumers, the UI, CLI, etc. to filter
service instances that are set in k8s. The Consul UI (in Consul 1.2.3 and later)
will read this value to show a Kubernetes icon next to all externally
registered services from Kubernetes.
Additional metadata can be specified using annotations. The "KEY" below can be set to any key. This allows setting multiple meta values:
kind: Service apiVersion: v1 metadata: name: my-service annotations: 'consul.hashicorp.com/service-meta-KEY': 'value'
»Consul Enterprise Namespaces
Consul Enterprise supports Consul namespaces. These can be used when syncing from Kubernetes to Consul (although not vice-versa).
There are three options available:
Single Destination Namespace – Sync all Kubernetes services, regardless of namespace, into the same Consul namespace.
This can be configured with:
global: enableConsulNamespaces: true syncCatalog: enabled: true consulNamespaces: consulDestinationNamespace: 'my-consul-ns'
Mirror Namespaces - Each Kubernetes service will be synced to a Consul namespace with the same name as its Kubernetes namespace.
For example, service
fooin Kubernetes namespace
ns-1will be synced to the Consul namespace
ns-1. If a mirrored namespace does not exist in Consul, it will be created.
This can be configured with:
global: enableConsulNamespaces: true syncCatalog: enabled: true consulNamespaces: mirroringK8S: true addK8SNamespaceSuffix: false
Mirror Namespaces With Prefix - Each Kubernetes service will be synced to a Consul namespace with the same name as its Kubernetes namespace with a prefix. For example, given a prefix
fooin Kubernetes namespace
ns-1will be synced to the Consul namespace
This can be configured with:
global: enableConsulNamespaces: true syncCatalog: enabled: true consulNamespaces: mirroringK8S: true mirroringK8SPrefix: 'k8s-' addK8SNamespaceSuffix: false
Note that in both mirroring examples we're setting
addK8SNamespaceSuffix: false. If set to
(the default), the Kubernetes namespace will be added as a suffix to each
Consul service name. For example Kubernetes service
foo in namespace
would be registered into Consul with the name
This is useful when syncing from multiple Kubernetes namespaces to
a single consul namespace but is likely something you'll want turned off
when mirroring namespaces since services won't overlap with services from
»Consul to Kubernetes
This syncs Consul services into first-class Kubernetes services.
The sync service will create an
Service for each Consul service. The "external name" will be
the Consul DNS name.
For example, given a Consul service
foo, a Kubernetes Service will be created
apiVersion: v1 kind: Service metadata: name: foo ... spec: externalName: foo.service.consul type: ExternalName
With Consul To Kubernetes syncing enabled, DNS requests of the form
will be serviced by Consul DNS. From a different Kubernetes namespace than where Consul
is deployed, the DNS request would need to be
Note: Consul to Kubernetes syncing isn't required if you've enabled Consul DNS on Kubernetes
and all you need to do is address services in the form
<consul-service-name>.service.consul, i.e. you don't need Kubernetes
Service objects created.
Requires Consul DNS via CoreDNS in Kubernetes: This feature requires that
Consul DNS is configured within Kubernetes.
is required (instead of kube-dns) to resolve an
issue with resolving
externalName services pointing to custom domains.
All Consul services visible to the sync process based on its given ACL token will be synced to Kubernetes.
There is no way to change this behavior per service. For the opposite sync direction (Kubernetes to Consul), you can use Kubernetes annotations to disable a sync per service. This is not currently possible for Consul to Kubernetes sync and the ACL token must be used to limit what services are synced.
In the future, we hope to support per-service configuration.
When a Consul service is synced to Kubernetes, the name of the Kubernetes service will exactly match the name of the Consul service.
To change this default exact match behavior, it is possible to specify a
prefix to be added to service names within Kubernetes by using the
-k8s-service-prefix flag. This can also be specified in the Helm
If a conflicting service is found, the service will not be synced. This does not match the Kubernetes to Consul behavior, but given the current implementation we must do this because Kubernetes can't mix both CNAME and Endpoint-based services.
»Kubernetes Service Labels and Annotations
Any Consul services synced to Kubernetes will be labeled and annotated.
consul.hashicorp.com/synced will be set to "true" to note
that this is a synced service from Consul.
Additionally, a label
consul=true will be specified so that label selectors
can be used with
kubectl and other tooling to easily filter all Consul-synced