Getting Started¶
Obtaining Access
Get in touch with Seldon for a trial.
Installation Types¶
General production installation. Modular installation into an existing cluster with instructions on how to leverage Seldon Enterprise Platform integrations. This is the recommended installation for most production setups.
Component integrations¶
Below are the integrations that Seldon Enterprise Platform supports, together with the versions currently supported for each of the components.
Component |
Supported Versions |
Integration level |
|---|---|---|
Seldon Core v1 |
1.16.0 and above |
Required |
Seldon Core v2 |
2.5.0 and above |
Required |
Ingress controller (e.g. Istio, NGINX, …) * |
Istio: 1.17.1+ |
Required |
PostgreSQL (Model Metadata Storage) |
9.6 and above |
Required |
Prometheus |
2.18.x and above |
Required |
OIDC provider (e.g. Keycloak or Dex) |
Keycloak 10.x+ |
Required |
Kafka |
3.3.1 and above |
Required |
Elasticsearch |
6.x and above |
Recommended |
ArgoCD |
2.0.x and above |
Recommended |
Argo Workflows |
3.0.x … 3.4.6 |
Optional |
Fluentd (or equivalent ELK log collection) |
6.x and above |
Optional |
Knative (Eventing, Serving) |
1.2.0 … 1.10.0 |
Optional |
Compatibility changes between versions are described in the upgrading notes.
Note
See discussion about which Ingress controller to use here.
Pre-requisites¶
Dependency |
Version |
Notes |
|---|---|---|
Kubernetes |
1.23 … 1.26 |
Seldon Enterprise Platform is intended for Kubernetes versions greater than or equal to 1.23 and less than or equal to 1.26. |
Git Server |
n/a |
To use GitOps, HTTPS access to a git server is required. GitOps is optional, but suggested. |
kubectl |
within one minor version difference of your cluster |
|
helm |
3.0.0 or greater is required |
Cluster Requirements¶
Seldon recommends the following minimum resource requirements for clusters running the full Seldon Enterprise Platform ecosystem. Depending on your usage, you may require more resources.
Cloud Provider |
Requirements |
Notes |
|---|---|---|
GCP |
3 x e2-standard-8 |
This is 24 vCPUs and 96GB RAM. Details. |
AWS |
6 x t2.xlarge |
This is 24 vCPUs and 96GB RAM. Details. |
Azure |
3 x D8as or D8ads |
This is 24 vCPUs and 96GB RAM. Details. |
Warning
Fewer and larger worker nodes are suggested as this requires less resource for the control plane. On EKS in particular increasing resources for the control plane has to be done explicitly. Too few nodes can also lead to problems on EKS with pods per node.