Jump to content

Connect Leeroopedia MCP: Equip your AI agents to search best practices, build plans, verify code, diagnose failures, and look up hyperparameter defaults.

Principle:Kserve Kserve Progressive Traffic Shifting

From Leeroopedia
Knowledge Sources
Domains MLOps, Deployment_Strategy, Traffic_Management
Last Updated 2026-02-13 00:00 GMT

Overview

A gradual rollout technique that incrementally increases the traffic percentage to a canary model revision through successive updates to the canaryTrafficPercent field.

Description

Progressive Traffic Shifting is the iterative process of increasing the canary traffic percentage after each validation checkpoint. Instead of an all-or-nothing switch, traffic gradually moves from the stable version to the canary (e.g., 10% → 20% → 50% → 100%).

Each increment is applied by updating the InferenceService spec with a higher canaryTrafficPercent value. The Knative service reconciler recalculates the traffic targets accordingly.

Usage

Use this between canary validation steps. After confirming the canary performs well at the current traffic level, increase the percentage. Common progression patterns are:

  • Conservative: 5% → 10% → 25% → 50% → 100%
  • Standard: 10% → 25% → 50% → 100%
  • Aggressive: 10% → 50% → 100%

Theoretical Basis

# Progressive shifting model (NOT implementation code)
For each increment:
  1. Update canaryTrafficPercent to new value P
  2. Knative reconciler recalculates:
     - LatestRevision.Percent = P
     - PreviousRevision.Percent = 100 - P
  3. Monitor canary metrics at new traffic level
  4. If acceptable → next increment
  5. If degraded → rollback to P = 0

Related Pages

Implemented By

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment