SLO priority levels

[emschwartz]

@keturiosakys and I talked to someone whose company has multiple SLOs that are grouped into different priority levels based on which part of the service they cover. Alerts for L1 functionality will wake engineers up in the middle of the night, while L2 and L3 might not.

This seems like a potentially interesting additional dimension for SLOs, but I'm not sure yet how we should think about this, or if it's the type of feature we should add to autometrics.

Some potential uses for this:

• Having the dashboard show groups of SLOs based on their priority
• Adjusting the alert severity / labels based on this priority
• Maybe the priority carries with it a default setting for the desired success rate and latency?

You could maybe argue that this is similar to Metric groups but it would be at a higher level than SLOs, rather than being a subset of functions in an SLO.

Thoughts