Behaviour when complete service outage

[mihaidraghici98]

Hello!

Does Sloth have a mechanism to include total service outages?

For example in a K8s deployment the pods might not exist at all or be excluded from the load balancer due to livenessProbe readinessProbe failing, thus the availability for the duration of incident being 0%.

I was trying to include such mechanism in the sli.events queries, but the 0% availability will be propagated to all windows and mess the SLI burn ratio (e.g. windows for 30d will be 0%, meaning a huge burn ratio).

Thanks!

[slok]

Hi @mihaidraghici98!

Sloth doesn't support that :/

I would suggest creating another SLO based on the Prometheus up metric. Even more, maybe you could play with the Prometheus target SLI plugin.

Best,