We use analytics and cookies to understand site traffic. Information about your use of our site is shared with Google for that purpose.You can read our privacy policies and terms of use etc by clicking here.
Namespace Setup
Setting up namespaces
Namespaces can have various features. A basic configuration can be done first. This is enough to use many features. Further features such as gitops can be added later.
Basic Namespace Setup
By default Seldon Deploy does not show any namespaces that are not explicitly labelled.
You can make a namespace globally available to all users by providing the following label to the namespace:
seldon.restricted: "false"This is the minimum needed to use a namespace.
Namespace Filtering
If the label seldon.restricted is set to “true” then visibility of the namespace can be narrowed by group or by user.
For details of this, see Namespace Visibility
Batch Jobs Permissions
Batch jobs use argo as a workflow engine. Argo needs a service account in the namespace and this has to be given permissions to create necessary resources. This is explained in the Argo section
Object Storage Secrets
The seldon deploy demos use models hosted in public buckets. To use private buckets, a secret containing environment variables should be setup in the namespace. An example is provided in the Argo section.
Models should reference the secret in the envFrom section of the wizard.
Gitops
A GitOps setup utilizes argocd to manage namespace contents in git. This requires setup for git, argocd and per-namespace. Each of these elements is covered in the gitops section
Multi-tenant Permissions (Only Used for Multi-tenant Installations)
In a multi-tenant setup, deploy is not automatically set up with permissions to work with multiple namespaces. Each namespace then needs further permissions. Details on this are in the multitenant section
Seldon-logs namespace (a special namespace)
The seldon-logs namespace is used for request logging. It is a special namespace and its setup is covered in the request logging section. It can be setup as a gitops namespace but this is optional. ML models do not go in this namespace.
Last modified April 20, 2021