FCIA experts took a deep dive on the role Fibre Channel plays in Kubernetes environments at our recent live webcast “Kubernetes and Fibre Channel: A Compelling Case.” Scott Sinclair, Practice Director and Senior Analyst at ESG set the stage with research on state of data storage, cloud, and the impact of containers. If you missed the live event, you can watch it on-demand at our FCIA YouTube Channel.
Our audience asked several interesting questions during the live event. As promised, here are answers to them all.
Q. What are the specific best practices for FC SAN fabric zoning when used with a Kubernetes cluster?
A. Peer zoning would be the perfect solution for zoning nodes eligible to host a particular containerized application with its corresponding storage array. This would allow for optimal container placement and migration while providing isolation between hosts.
Q. As a user, do I care how much persistent volumes were defined when I create my PVCs? What I mean by that is, if my storage admin has configured a total capacity of 20TB of PV, can I use 25TB as PVCs?
A. If an administrator has statically provisioned a persistent volume as 20TB then when that PV is bound to a persistent volume claim only 20TB will be available. The other option is to use dynamic provisioning with a CSI driver that supports volume expansion so that if additional capacity it needed after initial creation that volume can be grown. If configured, an alternate option could be to leverage the Kubernetes LimitRange construct.
Q. What about object storage support within CSI?
A. Object Storage is generally directly accessed via S3. There is a proposal from Red Hat to make ObjectBucketClaims and ObjectBuckets a standard feature of Kubernetes. It is currently limited to Red Hats products.
Q. Are specific HBA models and drivers required for Kubernetes persistent storage?
A. While specific HBA model number support for Kubernetes persistent storage should be verified with the FC HBA vendor, in general – most modern FC HBAs running the latest drivers and firmware are compatible with Kubernetes. It is however recommended that customers leverage the higher performance of 32G/64G FC HBAs and the low latency of FC-NVMe technology to drive efficiencies into their stateful workloads hosted on Kubernetes.
Q. Does CSI help with data migration from the public cloud to private?
A. There is work being done within the community to enable replication. Currently most CSI drivers do not have replication as a feature, which would help with migration between clouds.
Q. Do I need to purchase new FC hardware (HBAs, switches, storage…) for use with container persistent storage?
A. In most cases, customers would be able to leverage their existing Fibre Channel infrastructure as they modernize and containerize their application onto a Kubernetes platform. While most FC HBAs and FC switches already support and are compatible with Kubernetes persistent storage, it is imperative that you evaluate the Storage Array’s Kubernetes integration by verifying the availability of a CSI driver and its capabilities. Most modern FC/FC-NVMe storage arrays do provide a CSI driver with a robust set of capabilities.
