Guest blog author Sidharth Swami wrote this article. He is a Sr. Technical Account Manager here at VMware. As such Sidharth spends a lot of time with customers in real-world environments. Thank you, Sidharth, for sharing your time and expertise regarding build-your-own (BYO) vSAN clusters.
vSAN Turn-key Appliances and vSAN ReadyNodes are Recommended
One of the many benefits of deploying HCI powered by VMware vSAN is the wide variety of hardware platforms that can be used. Chances are your current x86 server vendor that sells hardware that is compatible with vSphere and vSAN. There is no need to switch hardware vendors to implement HCI (unless you want to).
The two primary deployment models are vSAN ReadyNodes and jointly engineered appliance solutions such as Dell EMC VxRail. vSAN ReadyNodes are available from all the leading x86 server vendors. These servers have been pre-configured, tested, and certified for vSAN. VxRail Appliances deliver a tightly-integrated, turnkey experience from initial deployment through “Day 2” lifecycle management.
Pre-configured vSAN appliances and vSAN ReadyNodes are the recommended platforms for deploying HCI with vSAN.
A third option is to build your own using hardware components certified for vSAN on the VMware vSAN Compatibility Guide. The BYO approach enables you to customize your HCI hardware to meet specific requirements. As with any customized solution, it also introduces additional complexity in terms of verifying hardware compatibility and support.
This article focuses on five important tips to ensure successful deployments of vSAN using the BYO approach.
1. Get the Right BYO Server Hardware
To promote existing server hardware equivalent to vSAN Ready Node it is important to check if it falls under VMware HCL (Hardware Compatibility List). The best place to confirm this compatibility is from VMware Compatibility Guide (VCG) and select "Build Your Own based on Certified Components." Here, one can choose certified components such as I/O controller, HDD, SSD compatible with vSAN/Hypervisor releases. A minimum of 3 nodes is required to set up a standard vSAN cluster. The same guidance applies to 2-node configurations as well. Refer to the vSAN 2 Node guide for more details.
2. Procure the Right Drives for Your Disk Groups
For each ESXi host (hybrid or all-flash) to be part of the vSAN cluster, it must have Cache Tier and Capacity Tier. Each host can have a maximum of 5 disk groups. Each disk group has one cache device and one to seven capacity devices. For hybrid configurations, the capacity tier has magnetic drives. All-flash configurations use flash devices in the cache and capacity tiers. These devices are available in various performance classes (Class B, C, D, E) based upon writes-per-second throughput. Before procuring new equipment, refer to VCG for supported configuration.
High-endurance flash devices are recommended for the cache tier. The quality of cache tier devices has an impact on the performance and endurance of the disk group. These more expensive cache devices can be balanced with lower-endurance drives in the capacity tier to balance cost and performance. Refer to SSD Endurance for vSAN.
3. Confirm Your Controllers Have the Right Firmware
VMware recommends using compatible controller firmware across a vSAN cluster. To identify supported controller firmware, details like VID (Vendor ID), DID (Device ID), SVID (Sub-Vendor ID), SSDIC (Sub-Device ID) are required. Refer to VMware KB 1027206 and use the VMware Compatibility Guide (VCG) to verify this information.
4. Don't Cut Corners with the vSAN Network
Use high-quality networking gear with high-performance features. Use a minimum of 10Gb for the vSAN network. 25Gb is recommended. Use switches that have sufficient buffer capacity. Make sure the NICs in your servers are also enterprise-class. Remember a chain is only as strong as its weakest link. All components of your storage network infrastructure must be up to the task of handling storage traffic. See this VMworld 2019 session video for more vSAN networking guidance.
5. Pay Attention to vSAN Health
vSAN has built-in health checks to verify its overall health and compare if the configurations are aligned to the VMware best practices. Always verify all health checks are “green” (no warnings or errors) before deploying production workloads to a vSAN cluster. Customers who enroll in CEIP (Customer Experience Improvement Program) receive dynamic updates for vSAN health and recommendations from VMware Analytic Cloud (VAC). Refer to Working with vSAN Health Checks.
Bonus Tip: ESXi Coredump and Scratch Partition Does Not Belong on vSAN
VMware does not support configuring scratch logs on vSAN datastore as it may cause some issues. The best way to configure a scratch log is to have it on a separate VMFS datastore or a SYSLOG server. Refer to VMware KB 2074026.
I'll say again that appliances such as Dell EMC VxRail and vSAN ReadyNodes are the recommended hardware platforms for HCI powered by vSAN. If you choose the BYO method, hopefully, these tips are helpful.