VMware Aria Operations Super Metric data contains gaps

Products

• VMware Aria Suite

Issue/Introduction

This article explains why gaps may appear in Super Metric data in VMware Aria Operations (formerly known as vRealize Operations).

Symptoms:

• Gaps of more than one collection cycle appearing in Super Metric data

Cause

When creating a Super Metric formula there are two options for specifying the metric key. 

• One option is to use the "metric" keyword, for example:

${this, metric=xyz|xyz}

Using the "metric" keyword means that VMware Aria Operations will find the exact match for metric key. 
As metric keys are unique, this keyword will never return more than one metric. (e.g. cpu|usage)

• The other option available is to use the "attribute" keyword, for example:

${this, attribute=xyz|xyz}

Using the "attribute" keyword means that VMware Aria Operations will find the exact match for metric key and also all of the matching instanced metric keys for the specified metric key. (e.g. cpu:1|usage, cpu:2|usage, etc.) 
In this case we potentially have a collection of metrics, so in order to do numeric operations with this collection VMware Aria Operations will need to convert it into single number. 
This conversion is done by aggregation functions (sum, max, avg, etc.).

Resolution

Avoid overuse of the "attribute" keyword when creating a Super Metric formula.
Instead use the "metric" keyword wherever possible.

Additional Information

Impact/Risks:

Performance comparison between "metric" and "attribute" keywords:

The "metric" keyword uses an exact match of the metric key, while the "attribute" keyword uses regex matching. 
Regex matching for the "attribute" keyword is much slower than exact match used by the "metric" keyword.
Overuse of "attribute" keyword in Super Metrics can potentially cause performance issues on the VMware Aria Operations cluster by degrading database performance and response time. 
This can then potentially create gaps in all Super Metric calculations, periodically missing several points in time.