Thanks Jason, Sharlene, Oliver & Preacta for that awesome meetup. I was thinking to myself it would be useful to try and capture some of the questions that were asked during. So, in no particular order....
During discussions about multi-cluster management, several key points emerged:
Someone from Redhat initiated a discussion about multi-cluster service management and Argo integration accross multiple large clusters. In general the conversation covered various multi-cloud/multi-cluster solutions:
-
The Cluster API (part of CNCF's Cluster Lifecycle SIG) has implementations across major cloud providers:
-
ByteDance (TikTok's parent company) recently contributed KubeAdmiral, their open-source multi-cluster solution, as mentioned in CNCF blog posts.
Regarding Argo Rollouts, two main topics were discussed:
-
Complex pre/post evaluation rules:
- Question focused on handling complex evaluations involving multiple data points converging to a single metric/decision
- Solution identified: Prometheus Recording Rules can handle complex PromQL queries that need to be expressed as single query calls
-
Traffic shifting and progressive rollouts:
- CNCF OpenFeature project was tangentally referenced in this context
On metrics collection, discussion covered Argo components' observability:
- Argo emits standard cloud-native service metrics
- Notable trend: Prometheus and similar tools are increasingly adopting OpenTelemetry standards