vSAN performance diagnostics provides feedback on how to extract the best performance in a given vSAN cluster. It consumes the data available via the vSAN performance service in vSAN health and provides details on performance issues seen in the VSAN cluster. To use vSAN performance diagnostics, you must join the Customer Experience Improvement Program (CEIP), and enable vSAN Performance Service (Cluster -> Configure -> vSAN -> Services -> Performance Service).

 

vSAN performance diagnostics requires the following user inputs :

1. Goal: This defines the context or user perspective of the performance evaluation. You should choose the desired evaluation type; that is, whether it is for maximum I/Os per second (Max IOPS), maximum throughput (Max Throughput), or minimum latency (Min Latency). Please remember that these evaluations may yield different results. As an example, the requirements for maximum IOPS are different than the requirements for minimum latency. The goal may be set to “Max IOPS”, “Max Throughput”, or “Min Latency”. The diagnostics evaluation will assume the selected objective for the analysis.
2. Time Range: This specifies the time duration for which vSAN performance diagnostics is required. The default time range is the last 1 hour. You can also increase it to the last 24 hours, or choose a custom time interval during the last 90 days. In case a vSAN proactive benchmark has been experimented with on this cluster, the time ranges for the proactive experiments will appear in the drop-down menu. Similarly, when HCIBench v1.6.2 is used to benchmark the cluster, the time range during which HCIBench was run will appear in the drop-down menu. If you are running some other benchmark, please make sure to run the benchmark for at least one hour, and then choose the start time of the benchmark and the duration (number of hours for which the benchmark was run). Performance diagnostics will not report any performance issues for an idle cluster. Instead, an issue mentioning: “No benchmark is seen to be running during the queried period” will be listed for an idle cluster.

After you click the Submit button, performance data for the specified time interval is transmitted to the analytics backend system. All data is obfuscated, compressed, and transmitted over the HTTPS protocol. After the data has been processed, the screen returns with a list of issues (or a message that no issues are found), and an Ask VMware link that documents how to find a solution for the issue. With each issue, there is an issue description that describes what the issue is about in a small paragraph. Each issue is accompanied with supporting data, which helps you to understand and debug the performance issue. The supporting data is usually in the form of performance graphs showing the entity, and metrics which explain the root cause behind why a performance issue was triggered. The graph shows data along with a threshold that indicates at which points in time the threshold was not met.

 

Possible issues, listed performance graphs, and possible remedies include:

• KB 2150014: One or more diskgroups(s) are not in active use
• KB 2150015: One or more disk(s) are not active
• KB 2150018: The increase in latency in the vSAN stack might be beyond expected limits
• KB 2150012 : vSAN is experiencing congestion in one or more disk group(s)
• KB 2150019: The size of IOs may not be optimal to achieve the desired goal
• KB 2150017: The outstanding IOs for the benchmark may not be optimal to achieve the desired goal
• KB 2150020: The vSAN cluster is seeing recovery traffic
• KB 2150011: The vSAN cache may not be sized correctly
• KB 2150026: No benchmark seems to have run during the selected time period.
• KB 2150181: There are errors seen in one or more physical NICs on one or more hosts
• KB 2150182: There are losses seen in the TCP layer at one or more hosts