etcd-io/etcd

[test] support cluster-api style rolling replacement upgrade

Open

#20804 opened on Oct 15, 2025

View on GitHub
 (8 comments) (0 reactions) (1 assignee)Go (51,701 stars) (10,352 forks)batch import
area/testinghelp wantedstage/trackedtype/feature

Description

What would you like to be added?

Currently our e2e framework supports in-place upgrade. We just replace the binary, and reuse the existing data.

https://github.com/etcd-io/etcd/blob/30452ab110473eda5f60bdb301c046194a154e91/tests/framework/e2e/downgrade.go#L131

However, the cluster-api/Kubeadm upgrades Kubernetes cluster in a rolling replacement approach,

  • add a new learner first;
  • promote it later
  • remove an old node
  • repeat the above steps until all nodes are replaced.

The key difference is the new member(learner) gets started without local data, so it may receive a snapshot from the leader.

We should update our e2e test framework to support both upgrade approaches,

  • rolling in-place upgrade
  • rolling replacement upgrade

Why is this needed?

To improve test coverage.

Contributor guide

Tech stack
go
Domain
testing
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
3-5 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
Go programmingetcd e2e test frameworkrolling upgrade concepts
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
35
Research direction
Study the existing e2e upgrade test framework in tests/framework/e2e/downgrade.go (linked). Understand the current in place upgrade flow. Research cluster api's rolling replacement approach for Kubernetes upgrades. Then implement a similar pattern for etcd, adding a new learner, promoting it, and removing old nodes.