Keynote: Pushing the Limits: Scaling ArgoCD’s Repo-Server To Handle 30...
summary
Expedia’s platform migrates to ArgoCD and scales the repo-server to generate manifests for 33,000 applications built from over 4,000 Helm charts; a custom Golang harness and 11 clusters plus ~100 vclusters simulate load. Observed repo-server pods spike to ~32 GB memory and 16 CPU cores when Helm runs concurrently, so monitoring uses DataDog and logs use Splunk. Stabilization uses parallelism=250, reconcile=240s, cache=24h, initial 8CPU/10GB up to 16CPU, and HPA 10→40; outcome: system stabilizes but requires tuning per workload.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:26Guest Introductions
- 1:02Agenda
- 2:43Repo
- 4:01HighLevel Design
- 5:11Summary
- 5:27Experiments Setup
- 7:49Request Test
- 8:24Key Algoity Settings
- 9:11Extra Tips
related talks
18mUnveil the Secret Ingredients for Argo CD in the Enterprise-Scale
26mGitOps Me Some of That! Managing Hundreds of Clusters with Argo CD
29mMastering Efficiency in Argo CD: Scaling Smarter, Not Costlier
23mArgo at Scale: Navigating Complex Multi-Dimensional Deployments Across H...
34mInfrastructure at Scale Using ArgoCD
29mTuning Argo Rollouts for Thousands of Workloads
15mA Quantitative Study on Argo Scalability
24mScaling Argo CD: From Symptoms To Solution