As organizations move from "GitOps experimentation" to "GitOps at scale," many hit a performance wall. Whether it’s an Application Controller struggling with OOM kills on 50,000 resources, a Repo Server buckling under "thundering herd" monorepo commits, or a GUI that lags under the weight of 5,000+ applications, scaling Argo CD requires more than just adding more RAM.
This session deconstructs the Argo CD scaling puzzle, piece by piece. We start by examining the internal plumbing of the Application Controller, demonstrating how to tune for vertical scaling by demonstrating how resource inclusions/exclusions, respectRBAC and Go runtime tuning can be configured to reduce resource usage and prevent OOM kills in high-density environments.
We will compare Dynamic Cluster Sharding (using consistent hashing) against the Argo CD Agent (Managed vs. Autonomous modes) for multi-cluster topologies. Finally, we will take a look at the future road map for improvements in the GUI and the progress towards Application-based sharding, which promises to decouple the Live State Cache from cluster boundaries and unlock true horizontal scalability for the next generation of GitOps.