Skip to main content

Workload identity and resolution

Otterize supports two approaches for specifying workload identity in Kubernetes: explicit and implicit.

Explicitly specifying Kind​

This approach requires specifying both the name, possibly with a namespace, and the kind of the workload in the intent, ensuring precise identification.

   apiVersion: k8s.otterize.com/v2beta1
   kind: ClientIntents
   metadata:
     name: client
     namespace: otterize-tutorial-istio-mapping
   spec:
     workload:
       name: client
       kind: Deployment
     targets:
       - kubernetes:
           name: server.example-ns
           kind: Service
       # This is equivalent to the above target
       - service:
           name: server.example-ns

In the YAML above, the workload client is a Deployment in the namespace otterize-tutorial-istio-mapping, and it intends to call the workload server, which is a Service, in the namespace example-ns.

Kubernetes workload identity resolution​

How do Otterize operators decide what is the identity of the workload that runs within the pod? The algorithm is as follows:

  1. If the pod has an intents.otterize.com/workload-name annotation, its value is used as the workload name. (You can change which annotation is used by setting global.workloadNameOverrideAnnotationName — see the docs.) This allows developers and automations to explicitly name workloads, if needed. The value must not contain a period . as a period is used to separate workload name and namespace, when the workload is from a different namespace: server.namespace.
  2. If there is no intents.otterize.com/workload-name annotation, a recursive look-up is performed for the Kubernetes resource owner of the pod, until the root resource is reached, and its name is used as the workload name. For example, if you have a Deployment named checkoutservice, which then creates and owns a ReplicaSet, which then creates and owns a Pod, then the workload name for that pod is checkoutservice - same as the name of the Deployment. This is intended to capture the likely-more-meaningful "human name" of the workload. If the resulting workload name contains a period ., it is replaced with an underscore _. Periods are used in workload names to denote namespaces, e.g. server.namespace.