Cost-Efficient Multi-Cluster Monitoring with Prometheus, Grafana & Linkerd - Carolin Dohmen, BWI

Cost-Efficient Multi-Cluster Monitoring with Prometheus, Grafana & Linkerd - Carolin Dohmen, BWI
Finding the right observability solution can be tough, especially in the context of diverse multi-cluster or even multi-cloud ecosystems. Cloud native monitoring tools like Prometheus provide a good foundation for a unified observability stack across clusters. But how do you stay on top of cluster and application metrics from a variety of source clusters? And what about secure data transfer, high egress costs, and expensive data redundancy? During this session, Carolin Dohmen will discuss how she and her former team at finleap connect, a German fintech company, built a decentralized monitoring solution based on Prometheus, Grafana, and Linkerd's multi-cluster service mirroring capability.

[@relo4727]

Very interesting and great presentation. I'm definitely going to look into this.

[@hz8711]

Hey, this is a pretty cool idea - central Grafana for multiple Prometheus data sources. Why are you so focused on Grafana's high availability, are you using it for alerts?

[@CarolinDohmen]

Hi! We needed to make sure that our central cluster is highly available, because we were running multiple services there, which are used by all the other workload clusters. So the main point was not to ensure that Grafana itself is HA, but rather the whole central services cluster. For alerting we used Alertmanager with a similar setup, so that all Prometheus instances push alerts to a central Alertmanager deployment.

[@larsvontrierpung9337]

mimir