Warewulf User Group | ZFS Storage with Warewulf + Project Updates скачать в хорошем качестве

Warewulf User Group | ZFS Storage with Warewulf + Project Updates

In this session, Josh Burks, HPC Systems Analyst at Arizona State University, walks through how ASU uses ZFS with Warewulf to build flexible, memory-efficient, stateless-OS / stateful-data hybrid nodes across four clusters and roughly 1,000 nodes — including their new Kubernetes-based AI cluster running Intel Gaudi 2 accelerators. Josh covers why ZFS pairs naturally with stateless provisioning, how to configure ZFS pools dynamically via wwwinit.d scripts, and real-world use cases including persistent Hugging Face model caches, Kubernetes per-node credentials, and software RAID for storage nodes without hardware RAID controllers. Jonathon Anderson (TSC Chair) opens with project updates, including the TLS/HTTPS upgrade to the Warewulf daemon, a server code refactor that bumped the release from v4.6.6 to v4.7.0, an updated Golang version policy, and closure of all known Dependabot and Snyk vulnerabilities. Topics covered: Warewulf 4.7.0 release preview: TLS support, server rework, Golang 1.25 ZFS overview: pools, datasets, compression, automatic mounting, DKMS Stateless OS + stateful data hybrid architecture Provision-to-disk with ZFS instead of XFS/EXT4 partitions Building ZFS kernel modules inside a Warewulf container image Kubernetes + Warewulf: persistent node identity across reboots CDDL vs GPL2 licensing and what it means for your deployment Resources mentioned: Josh's example ZFS scripts: https://github.com/jeburks2/warewulf-...