Goal

Release the agent-based multi-cluster Argo CD as Generally Available

Background

• Customers have moved past running single instances of Argo CD on single clusters
• Scale is the major blocker for large production implementations, across many facets - performance, observability, instance management, and security
• Customers and internal teams have expressed interest in Akuity’s Agent model
• Edge and far-edge customers require the most minimal possible footprint, today you need to run all of the parts of Argo CD in the same cluster for it to work - if we could remove this interdependency it would open the door to breaking up the parts across clusters and reducing the installation footprint even further

Business goal

• Provide a multi-cluster capability in OpenShift GitOps that addresses scaling issues across multiple areas and prepares customers for much larger and more complex production deployments than they’re able to do today

Dependencies

• GitOps team to lead the majority of the work, and to lead the work with the upstream community on adding this functionality to Argo CD
• ACM team

Customer Need

Specific customer need (external/internal) that this outcome will solve:

• Scaling issues in a variety of areas
• Scale in performance of the Argo CD components under load, but also scale in regards to observability of Argo CD instances, managing those instances across clusters, and keeping everything secure with distributed instances or when large numbers of users need to access central instances

• Customers are expressing a preference for the pull model
• Security/trust issues that come with the push model

Customers that we are engaging for feedback:

• Ford
• Amadeus
• Lockheed
• Swiss Railways

Success Criteria

• There exists a central place that customers can view all of their Argo CD instances across clusters
• Argo CD instances and their configuration can be managed and lifecycled across clusters from a central location
• Access control for instances can be centrally managed and can utilise ACM policies

Expected Results (what, how, when)

• Performance improvements when running large volumes of applications (an increase in Argo CD’s current capacity/limits)

Links

Discussion doc: Multi-cluster GitOps and the Argo Agent proposal

Architectural overview: Work items breakdown

High level deck: Layered products in multi-cluster environments

Argo agent overview: Introducing argocd-agent

Scaling problems and the agent approach: Argo CD Scaling 

Argo agent upstream repo and architectural docs: argocd-agent

Jann's presentation of architecture (November 2023): Recording

Jann's presentation of architecture in cabal (July 2024): Recording

Metrics

Post Completion Review – Actual Results

TBD