The Indian Premier League is a unique sporting event for a dozen reasons. But for engineers in India, it’s one of a kind. Very few companies can boast of managing 30+ million concurrent users. Every year, this number grows. Last year, we witnessed ~60 million concurrent users. And things get bigger and larger every year.

Predicting the future is hard, especially with metrics-based monitoring systems, because metrics cardinality can snowball. This is important because it affects query performance adversely. Having visibility into what’s happening now and workflows to manage cardinality is crucial. Because the answers depend on the quality of questions, a system allows you to ask. The questions one may have is —

We have Carson Anderson, Sr. DevOps Engineer at Weave HQ, talking about how they implemented SLOs using Prometheus, what went wrong, and how they fixed it. This talk was given at "Last9 of Reliability" Discord community on 13th December. Talk Description: First thing's first: Yes, it really did take us 5 tries to implement our SLOs with Prometheus. While that may seem embarrassing, we are very happy to be able to share our SLO journey so that we can hopefully help you avoid the same mistakes.