---
title: "Discussion on \"Canary Releases for an ML Model on KServe: Replayed Traffic, Sidecar Metrics and Automatic Rollback\" | Hashnode"
description: "Discussion on \"Canary Releases for an ML Model on KServe: Replayed Traffic, Sidecar Metrics and Automatic Rollback\". Canary Releases for an ML Model on KServe: Replayed Traffic, Sidecar Metrics and Automatic Rollback"
url: "https://hashnode.com/posts/kserve-model-canary-deployment-automation/6ae3c650276c1175b0a3da04"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Canary Releases for an ML Model on KServe: Replayed Traffic, Sidecar Metrics and Automatic Rollback](https://georgeezejiofor.com/kserve-model-canary-deployment-automation?utm_source=hashnode&utm_medium=feed)

Canary Releases for an ML Model on KServe: Replayed Traffic, Sidecar Metrics and Automatic Rollback

Oct 29 · 23 min read · georgeezejiofor.com  
Read the article

-   [#kubernetes](https://hashnode.com/tag/kubernetes)
-   [#machine-learning](https://hashnode.com/tag/machine-learning)
-   [#mlops](https://hashnode.com/tag/mlops)
-   [#kserve](https://hashnode.com/tag/kserve)
-   [#kubeflow](https://hashnode.com/tag/kubeflow)
-   [#canary](https://hashnode.com/tag/canary)
-   [#canary-deployment](https://hashnode.com/tag/canary-deployment)
-   [#canary-release](https://hashnode.com/tag/canary-release)
-   [#deployment](https://hashnode.com/tag/deployment)
-   [#deployment-automation](https://hashnode.com/tag/deployment-automation)
-   [#deployment-strategies](https://hashnode.com/tag/deployment-strategies)
-   [#training](https://hashnode.com/tag/training)

## Written by

[George Ezejiofor](https://hashnode.com/@georgeezejiofor)

@georgeezejiofor · London, United Kingdom

  
[View profile](https://hashnode.com/@georgeezejiofor)

Director & AI/ML Platform / MLOps Engineer at GeoCloudTech | Kubestronaut | Kubernetes, DevSecOps and MLOps for regulated industries

I'm an AI/ML Platform and MLOps Engineer and a Kubestronaut, with 13+ years in IT and 5+ in DevOps and DevSecOps. Through my consultancy, GeoCloudTech, I design secure, highly available Kubernetes platforms for regulated industries, on cloud and on-premises. What I build: - Kubernetes platforms: RKE2, AKS, EKS, GKE - MLOps: Kubeflow, KServe, MLflow, NVIDIA GPU Operator - DevSecOps: Vault, Kyverno, KubeArmor, Trivy, signed supply chains with Tekton and Cosign - GitOps and automation: Argo CD, Flux, Terraform, Ansible - Storage and identity: Rook Ceph, CloudNativePG, Keycloak - Observability and networking: Prometheus, OpenTelemetry, Cilium, Istio Most of what I publish comes from my multi-datacentre homelab: RKE2 on Proxmox, run through GitOps. My rule: build it, break it, understand it, then harden it. Every article is a real build, traps included. Certifications: Kubestronaut (CKA, CKAD, CKS, KCNA, KCSA), HashiCorp Terraform Associate.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
