Built for One, Merged for Everyone: IPv6-Native HA Storage for a Tier-1 Telco — and What We Gave Back Upstream
When a tier-1 mobile network operator evaluated software-defined storage for its petabyte-scale cloud platform, the requirements were demanding: replicated block storage for both Kubernetes and bare-metal workloads, highly available multiple NFS exports — all running end-to-end on IPv6. This talk covers how we built that platform on a fully open source stack: LINSTOR and DRBD for replicated storage, CSI for Kubernetes, and Pacemaker-managed NFS with Virtual IP failover. Along the way, the IPv6 requirement exposed a gap in the ClusterLabs portblock resource agent — so we built IPv6 support and merged it upstream, alongside accelerating NFS-Ganesha integration in linstor-gateway. A case study in why upstreaming customer-driven fixes beats carrying vendor patches.
Speaker
-
Yusuf YildizLINBITWith a robust background spanning 15 years in system and storage management, Yusuf transitioned into the role of Solution Architect at LINBIT in 2009. Dedicated to customer success, he plays a hands-on role in product testing and development. As a foundational pillar of the architecture team, Yusuf provides critical support to both the sales and development departments, strategically directing product evolution.