Namespaces and groups for Kubernetes experiments¶

Kubernetes experiments are launched within a namespace, and are associated with a unique group within that namespace.

For example, consider the following invocation:

iter8 k launch -g hbin \
--set "tasks={http,assess}" \
--set http.url=http://httpbin.default/get \
--set assess.SLOs.upper.http/latency-mean=50

In the above invocation, the iter8 k launch command implicitly specifies the namespace as default, and explicitly specifies the group as hbin. If the group name is not specified explicitly, then it is set to default. The namespace can be specified explicitly using the -n or --namespace flags (see here).

The following example illustrates the relationship between namespaces, groups, and experiments.

.
├── namespace1
│   ├── group-a
│   │   └── experiment
│   ├── group-b
│   │   └── experiment
│   └── group-c
│       └── experiment
├── namespace2
│   ├── group-a
│   │   └── experiment
│   ├── group-b
│   │   └── experiment
│   └── group-c
│       └── experiment
└── namespace3
    └── group-x
        └── experiment

Use-cases¶

1. Run multiple experiments concurrently within a Kubernetes namespace by associating them with distinct groups. These experiments may be associated with the same app or with different apps.
2. Replace a currently running experiment in Kubernetes with a new one. When you invoke iter8 k launch, any previous experiment executions within the group is wiped out and replaced with a fresh experiment that starts to execute.

How groups work¶

Under the covers, Iter8 implements each experiment group as a Helm release and each new experiment run within the group as an update of that release.