vSAN Stretched Cluster Improvements in vSAN 8 U2
Key Points
vSAN 8 U2 has added several new capabilities and improvements, including:
- The Introduction of vSAN Max – Disaggregated scale out storage done the right way
- ESA Scalability Improvements – 500 VMs per host
- ESA Performance Improvements – Built for speed
- Interoperability Improvements – vSAN File Services and Cross-Cluster Support
- Prescriptive Disk Claiming – Faster and easier
- Health and Performance Monitoring enhancements – Top Contributors, NVMe Endurance Tracking and Skyline Health Enhancements
There are also improvements to vSAN Stretched Clusters in vSAN 8 U2; this post will walk you through them.
More straightforward configuration for Stretched and 2-node Clusters
Witness Traffic Separation for vSAN OSA and ESA
vSAN 8 U2 makes the configuration of stretched clusters and 2-node topologies easier than ever. Customers can now tag vSAN witness traffic in the witness host appliance through the VMkernel configuration settings for the virtual appliance. Tagging this traffic through the UI removes the extra steps and hassle needed to perform this tagging of witness host appliance traffic through a command line. The ability to tag a VMkernel interface will be available when creating a new VMkernel port and editing an existing VMkernel port.
This feature enhancement applies to OSA and ESA and is functional for both 2-node and stretched clusters.
New support of “Medium” sized witness host appliance in ESA
We’ve also improved the witness host appliance sizes available for the vSAN Express Storage Architecture (ESA) in vSAN 8 U2. In addition to the “large” sized witness host appliance, customers running these configurations can also choose a “medium” sized witness host appliance, which consumes 2vCPUs and 16GB of RAM but will support 500 VMs and 21,000 components. This new size increases flexibility and provides the opportunity to free up resources for other workloads.
Removal of the “None-Stretched Cluster” storage policy rule
vSAN 8 U2 also simplifies storage policy options for stretched clusters by removing the “None – Stretched cluster” option. The “None – Stretched cluster” policy was a policy placement rule for stretched clusters that did not replicate data for site-level resilience and placed object data across sites arbitrarily. It resulted in no defined site location of the objects and no defined location of the associated VM. Using this policy could result in potentially non-optimized I/O paths and the potential for objects being inaccessible in the event of planned or unplanned downtime. The rule provided no realistic benefit to a stretched cluster environment and was occasionally prone to user misinterpretation. Removing it simplifies the user experience and reduces the chance of errors. This change applies to both OSA and ESA.
Improved Detection of Performance Bottlenecks in Stretched Clusters
Recent versions of vSAN have made it much easier for our customers to identify the source of a performance bottleneck. The vSAN I/O Trip Analyzer is one of those helpful tools, and with vSAN 8 U2, we’ve enhanced its capabilities to perform an analysis on workloads running in a vSAN stretched cluster. Now, a user can quickly determine where the primary source of latency is occurring in a vSAN stretched cluster and latencies in other parts of the stack that may contribute to the overall latency seen by the VM.
vSAN 7 U3 introduced the VM I/O Trip analyzer to identify the potential point of contention when a VM is experiencing performance issues. It was a capable tool that could help diagnose VM-specific performance issues but could not schedule diagnostics runs. The I/O Trip Analyzer was limited to standard vSAN cluster topologies and did not support a stretched cluster configuration. Unfortunately, a stretched cluster configuration can be one of the most difficult for customers to diagnose the cause of an identified performance issue.
vSAN 8 U2 enables the VM I/O Trip Analyzer in stretched cluster and 2-node topologies. In these topologies, the I/O Trip Analyzer now has awareness and the ability to measure the time to process/transmit I/O to/from one site (or FD) to the other site (or FD). This process/transmit time is a critical aspect of identifying the source of latency, as the ISL may or may not be the primary cause of latency.
The tool will highlight any condition where it identifies ISL network latency exceeding 5ms and denote that visual connection link as red. The VM I/O Trip Analyzers highlights and callouts align with our product requirements and design guidelines for a stretched cluster environment.
This feature is available in vSAN OSA and ESA.
Summary
Our engineering teams have continued the impressive feat of enhancing and improving on already outstanding solutions. vSAN 8 U2 is no exception. It includes capabilities that make setting up, using, and troubleshooting easier. We are excited to hear about how you used these new features.
Useful links:
- Using the vSAN ESA in a stretched cluster topology: https://core.vmware.com/blog/using-vsan-esa-stretched-cluster-topology
- Performance with vSAN stretched clusters: https://core.vmware.com/blog/performance-vsan-stretched-clusters