Introduction
This section is for NDIP administrators. It covers the infrastructure NDIP runs on, the services that make up the platform, how to deploy them, and how to troubleshoot problems.
NDIP at a glance (October 2026)
NDIP's core services (Galaxy, Rucio, PostgreSQL, RabbitMQ, the Remote Data Broker, the NOVA dashboard and the ONCat proxy) run on two Kubernetes clusters managed through Rancher:
| Cluster | Environment | URL |
|---|---|---|
| mickey | test | https://ndip-test.ornl.gov |
| speedy | production | https://ndip.ornl.gov |
User jobs still run on Pulsar compute nodes (SNS analysis cluster, imaginex, OLCF Slate, NERSC Spin), which are deployed separately. Until the F5 load balancer and DNS cutover are done, production traffic reaches speedy through the old production VM, which now works only as a TLS-passthrough proxy.
Where to start:
- Infrastructure overview: what runs where, including what is still on VMs.
- Kubernetes clusters: access, namespaces, storage and limits on mickey and speedy.
- Deploying NDIP: the GitLab CI jobs for each component and environment, secrets, and charts.
- Troubleshooting guide: common problems on the clusters and the Pulsar hosts.
Pages describing the old VM/Ansible deployment of the core services have been updated or marked legacy. The
Ansible content that still applies covers the Pulsar hosts, galaxy.ornl.gov, the Rucio VM that serves
galaxy.ornl.gov, the docs server and the ingress-proxy VM.