mirror of
https://github.com/kubernetes-sigs/external-dns.git
synced 2025-10-31 18:50:59 +01:00
120 lines
5.7 KiB
Markdown
120 lines
5.7 KiB
Markdown
# Service source
|
|
|
|
The service source creates DNS entries based on `Service` resources.
|
|
|
|
## Filtering the Services considered
|
|
|
|
The `--service-type-filter` flag filters Service resources by their `spec.type`.
|
|
The flag may be specified multiple times to allow multiple service types.
|
|
|
|
This source supports the `--label-filter` flag, which filters Service resources
|
|
by a set of labels.
|
|
|
|
## Domain names
|
|
|
|
The domain names of the DNS entries created from a Service are sourced from the following places:
|
|
|
|
1. Adds the domain names from any `external-dns.alpha.kubernetes.io/hostname` and/or
|
|
`external-dns.alpha.kubernetes.io/internal-hostname` annotation.
|
|
This behavior is suppressed if the `--ignore-hostname-annotation` flag was specified.
|
|
|
|
2. If no DNS entries were produced for a Service by the previous steps
|
|
and the `--compatibility` flag was specified, then adds DNS entries per the
|
|
selected compatibility mode.
|
|
|
|
3. If no DNS entries were produced for a Service by the previous steps
|
|
or the `--combine-fqdn-annotation` flag was specified, then adds domain names
|
|
generated from any`--fqdn-template` flag.
|
|
|
|
### Domain names for headless service pods
|
|
|
|
If a headless Service (without an `external-dns.alpha.kubernetes.io/target` annotation) creates DNS entries with targets from
|
|
a Pod that has a non-empty `spec.hostname` field, additional DNS entries are created for that Pod, containing the targets from that Pod.
|
|
For each domain name created for the Service, the additional DNS entry for the Pod has that domain name prefixed with
|
|
the value of the Pod's `spec.hostname` field and a `.`.
|
|
|
|
## Targets
|
|
|
|
If the Service has an `external-dns.alpha.kubernetes.io/target` annotation, uses
|
|
the values from that. Otherwise, the targets of the DNS entries created from a service are sourced depending
|
|
on the Service's `spec.type`:
|
|
|
|
### LoadBalancer
|
|
|
|
1. If the hostname came from an `external-dns.alpha.kubernetes.io/internal-hostname` annotation, uses
|
|
the Service's `spec.clusterIP` field. If that field has the value `None`, does not generate
|
|
any targets for the hostname.
|
|
|
|
2. Otherwise, if the Service has one or more `spec.externalIPs`, uses the values in that field.
|
|
|
|
3. Otherwise, iterates over each `status.loadBalancer.ingress`, adding any non-empty `ip` and/or `hostname`.
|
|
|
|
If the `--resolve-service-load-balancer-hostname` flag was specified, any non-empty `hostname`
|
|
is queried through DNS and any resulting IP addresses are added instead.
|
|
A DNS query failure results in zero targets being added for that load balancer's ingress hostname.
|
|
|
|
### ClusterIP (headless)
|
|
|
|
Iterates over all of the Service's Endpoints's `subsets.addresses`.
|
|
If the Service's `spec.publishNotReadyAddresses` is `true` or the `--always-publish-not-ready-addresses` flag is specified,
|
|
also iterates over the Endpoints's `subsets.notReadyAddresses`.
|
|
|
|
1. If an address does not target a `Pod` that matches the Service's `spec.selector`, it is ignored.
|
|
|
|
2. If the target pod has an `external-dns.alpha.kubernetes.io/target` annotation, uses
|
|
the values from that.
|
|
|
|
3. Otherwise, if the Service has an `external-dns.alpha.kubernetes.io/endpoints-type: NodeExternalIP`
|
|
annotation, uses the addresses from the Pod's Node's `status.addresses` that are either of type
|
|
`ExternalIP` or IPv6 addresses of type `InternalIP`.
|
|
|
|
4. Otherwise, if the Service has an `external-dns.alpha.kubernetes.io/endpoints-type: HostIP` annotation
|
|
or the `--publish-host-ip` flag was specified, uses the Pod's `status.hostIP` field.
|
|
|
|
5. Otherwise uses the `ip` field of the address from the Endpoints.
|
|
|
|
### ClusterIP (not headless)
|
|
|
|
1. If the hostname came from an `external-dns.alpha.kubernetes.io/internal-hostname` annotation
|
|
or the `--publish-internal-services` flag was specified, uses the `spec.ClusterIP`.
|
|
|
|
2. Otherwise, does not create any targets.
|
|
|
|
### NodePort
|
|
|
|
If `spec.ExternalTrafficPolicy` is `Local`, iterates over each Node that both matches the Service's `spec.selector`
|
|
and has a `status.phase` of `Running`. Otherwise iterates over all Nodes, of any phase.
|
|
|
|
Iterates over each relevant Node's `status.addresses`:
|
|
|
|
1. If there is an `external-dns.alpha.kubernetes.io/access: public` annotation on the Service, uses both addresses with
|
|
a `type` of `ExternalIP` and IPv6 addresses with a `type` of `InternalIP`.
|
|
|
|
2. Otherwise, if there is an `external-dns.alpha.kubernetes.io/access: private` annotation on the Service, uses addresses with
|
|
a `type` of `InternalIP`.
|
|
|
|
3. Otherwise, if there is at least one address with a `type` of `ExternalIP`, uses both addresses with
|
|
a `type` of `ExternalIP` and IPv6 addresses with a `type` of `InternalIP`.
|
|
|
|
4. Otherwise, uses addresses with a `type` of `InternalIP`.
|
|
|
|
Also iterates over the Service's `spec.ports`, creating a SRV record for each port which has a `nodePort`.
|
|
The SRV record has a service of the Service's `name`, a protocol taken from the port's `protocol` field,
|
|
a priority of `0` and a weight of `50`.
|
|
In order for SRV records to be created, the `--managed-record-types` must have been specified, including `SRV`
|
|
as one of the values.
|
|
|
|
```console
|
|
external-dns ... --managed-record-types=A --managed-record-types=CNAME --managed-record-types=SRV
|
|
```
|
|
|
|
### ExternalName
|
|
|
|
1. If the Service has one or more `spec.externalIPs`, uses the values in that field.
|
|
2. Otherwise, creates a target with the value of the Service's `externalName` field.
|
|
|
|
## Endpoints Reconciliation
|
|
|
|
By default, ExternalDNS does not watch for endpoint changes and does not automatically reconcile DNS records as the endpoints, as matched by the Service's selector.
|
|
To enable reconcile on endpoints changes, you must specify the `--listen-endpoint-events` flag. However, be aware that this may increase the number of reconciliations performed by the controller, and the number of requests to the DNS provider.
|