kubernetes-sigs/cluster-api

CoreDNS pods running on the same Node after initial cluster creation

Open

#9,639 opened on Oct 30, 2023

View on GitHub
 (17 comments) (2 reactions) (0 assignees)Go (1,532 forks)auto 404
area/provider/control-plane-kubeadmhelp wantedkind/featurepriority/important-longtermtriage/accepted

Repository metrics

Stars
 (4,267 stars)
PR merge metrics
 (PR metrics pending)

Description

After a cluster with multiple control plane nodes is created all CoreDNS replicas are running on the same Node. Ideally they should be distributed across multiple nodes

Details

The kubeadm bootstrap provider is using kubeadm to deploy CoreDNS. It creates the CoreDNS deployment with preferredDuringSchedulingIgnoredDuringExecution. Because at the time of CoreDNS deployment there is only one Node all CoreDNS pods are scheduled on the same Node. There is no re-scheduling once additional CP nodes are added.

Related kubeadm issues, let's take a look!

Some idea on how to solve this

  1. Change preferredDuringSchedulingIgnoredDuringExecution to requiredDuringSchedulingIgnoredDuringExecution for multi-CP control planes (also change it back during scale down to 1 replica)
    1. we should check if replica is always the same number (e.g. 2)
  2. Delete CoreDNS Pods as soon as additional CP replicas join

Contributor guide

Research direction
Investigate how to modify the CoreDNS deployment to use requiredDuringSchedulingIgnoredDuringExecution for multi CP clusters, or delete CoreDNS pods when new CP nodes join. Examine the related kubeadm issues #1657 and #1931 for context and potential approaches.
Tech stack
go
Domain
backendinfrastructuredevops
Issue type
Bug
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.
1-2 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
KubernetesGokubeadm
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
30