NAME

Multi-cluster Kubernetes gateway proxies - One Determined master running tasks across multiple Kubernetes clusters through gateway proxies

SYNOPSIS

2024; Go, Kubernetes, Helm, Python

DESCRIPTION

Feature in the open-source Determined platform that lets a single master schedule tasks across multiple Kubernetes clusters and resource managers. Gateway resources deploy alongside each task's rank 0 pod, and task ports are translated to a configurable range of gateway listener ports so proxied traffic reaches tasks in remote clusters. Led the project at HPE through implementation, tests, and user docs, and spun out a developer tool that connects a local dev cluster to remote Kubernetes clusters.

HIGHLIGHTS

• Port translation from gateway listeners to proxied task ports, with a configurable and validated port range
• Gateway resources deployed only with rank 0 pods, and gateway startup hardened to run without passwordless sudo
• Multi-node, multi-port test experiments in CI, port registry tests, and task logs captured on failure
• Authored the user documentation for gateway setup
• Built a developer tool to connect a local dev cluster to remote Kubernetes clusters

SEE ALSO

* projects(7), all projects
* ari(7), the resume
* PDF Resume: Download Resume