I'm trying to understand the purpose and expected usage of the "Host CPU Usage Utilization" SLO template.
When creating this SLO, the default target is set to 98%. However, CPU utilization for healthy hosts is typically well below this value. As a result, the SLO remains in a failed/red state, even when the host is operating normally.
For example:
- Host CPU utilization is consistently around 5-12%
- SLO target is 98%
- SLO status remains red with significant error budget consumption
My expectation for a CPU utilization SLO would be:
- Good event: CPU utilization remains below the saturation threshold (e.g., <85% or <98%)
- Bad event: CPU utilization exceeds the saturation threshold
In other words, if a host's CPU utilization stays below 98% for the entire evaluation period, I would expect the SLO to be healthy (green), since the system is not saturated.
However, the current template appears to evaluate the raw utilization value against the target, which effectively means higher CPU utilization is considered better. This seems counterintuitive for a saturation-oriented metric.
Could someone help clarify:
- What is the intended use case for the Host CPU Usage Utilization SLO template?
- Is the template designed to measure CPU utilization itself rather than compliance with a saturation threshold?
- If the goal is to measure CPU saturation, would a custom DQL-based SLI be required to model "CPU utilization remains below X%" as the success condition?
- Are there recommended best practices from Dynatrace for creating infrastructure/resource utilization SLOs?
I may be misunderstanding the intent of this template and would appreciate any guidance from the Dynatrace team or community members.
Thank you!
