Introduction
This section provides detailed information about the services used within the NDIP platform, including instructions for their deployment and configuration.
The NDIP consists of two primary components:
- Core services: internal services the platform needs to operate. They run in namespace
ndipon the mickey (test) and speedy (prod) Kubernetes clusters. - Worker node services: services that run on the compute resources where users execute their jobs:
- Neutron Science analysis cluster and imaginex (Ansible)
- OLCF Frontier, running on the Slate Kubernetes cluster
- NERSC Perlmutter, running on the Spin Kubernetes cluster
note
For information about infrastructure supporting NDIP please refer to the infrastructure section. For how to deploy each service, see Deploying NDIP.
Currently in use:
| Name | Function | Runs on | Details |
|---|---|---|---|
| Galaxy | Main component of the NDIP platform | mickey / speedy | link |
| PostgreSQL (CloudNativePG) | Galaxy and Rucio databases | mickey / speedy | link |
| Pulsar | Accepts jobs from Galaxy and runs them on compute resources | compute hosts (Ansible), Slate, Spin | link |
| RabbitMQ | Message broker between Galaxy and Pulsar (and Celery, RDB) | mickey / speedy | link |
| Rucio | Manages NDIP data | mickey / speedy (Rucio VM for galaxy.ornl.gov) | link |
| Remote Data Broker | Moves data between worker nodes and Galaxy storage | mickey / speedy, compute hosts | link |
| NOVA Dashboard | Launcher for interactive NOVA tools, at /nova | mickey / speedy | link |
| ONCat Proxy | Authenticated access to ONCat, at /oncat | mickey / speedy | link |
| Apache Kafka | Stores data ingress requests coming from the ingress proxy | ingress-proxy VM | link |
| Ingress Proxy | Listens for ingress requests and sends them to Kafka | ingress-proxy VM | link |
| Prometheus stack | Monitoring and alerting; also gates job routing | monitoring cluster | link |
| kube-state-metrics | Workload and node state of mickey/speedy for Prometheus | mickey / speedy | link |
| Node Exporter | Exposes hardware/system metrics for Prometheus | VMs and compute hosts | link |
| Logging | Pod logs (Loki), ELK (in progress) | mickey / speedy | link |
| Documentation | This site, plus /nova/tutorial | docs VM | link |
| galaxy-mcp | MCP server for Galaxy, at /mcp | old prod VM (temporary) | link |
In the following sections, we provide details about these services.