Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

kubernetes: a single 1-vCPU worker capped at one node cannot fit the cert-manager install

Aperta
  • Qualcosa non va?
#4,774 0 commenti 1 reazione 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 2 giorni

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
3/5
Tempo stimato
Mezza giornata
Idoneità per principianti
65/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
helm, kubernetes
Ambito
devops

Direzione di ricerca

The issue is in the cert-manager addon chart deployment. Start by reviewing the CPU requests for all addons in the Helm values to confirm the overcommit on 1-vCPU nodes. The fix is either to add a validation warning in the chart when a 1-vCPU pool is used with addons enabled, or to document a minimum 2-vCPU requirement. Key files: packages/apps/kubernetes-nodes/values.yaml and the cert-manager chart values.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

triage/needs-triage

A tenant cluster whose only worker pool is one 1-vCPU node that cannot scale out (maxReplicas: 1) can't finish installing the cert-manager addon. The other addons already take that node's whole allocatable CPU. cert-manager's startupapicheck hook Job then has nowhere to run, the install times out, and the reinstall after remediation hits the same full node, so the HelmRelease never turns Ready.

1 vCPU is what a pool gets by default. The default instanceType is u1.medium (kubernetes-nodes values), and every *.nano, *.micro, *.small and *.medium type in packages/system/kubevirt-instancetypes has one vCPU. The default maxReplicas: 10 hides the problem, because the autoscaler adds a node for the Pending pod. It shows up once a pool is capped at one node.

The numbers, at d36784603:

  • The kubelet reservations for a 1-vCPU worker are 50m system plus 50m kube (_talosconfigtemplate.tpl), which leaves 900m allocatable.
  • helm template of the addon charts with the values a ComputePlane cluster passes them gives these per-node CPU requests: cilium agent, envoy and operator 300m, coredns 2 x 100m, kubevirt-csi-node 20m, ingress-nginx controller plus protobuf-exporter 200m and defaultbackend 10m, metrics-server 100m, cert-manager controller, webhook and cainjector 70m. That is 900m.
  • The startupapicheck Job asks for 10m more.

The pod only fits if it gets scheduled before the node fills up, and it then completes and gives the 10m back. So the result depends on install order. I saw this in the computeplane e2e suite, which ran one such worker and failed once and passed once on the same charts. The failing run's HelmRelease reported timeout waiting for: [Job/cozy-cert-manager/cert-manager-startupapicheck status: 'InProgress']. No FailedScheduling event was observed, since nothing inside the tenant cluster was collected (#4773). Pending is inferred from the arithmetic and from the cert-manager webhook answering the tenant apiserver for the whole remaining window. The suite now runs two workers.

A ComputePlane cluster always enables cert-manager and ingress-nginx, so it hits this with any single 1-vCPU pool. A plain Kubernetes cluster hits it once addons.certManager.enabled is on.

Possible fixes: have the chart reject or warn about a pool with maxReplicas: 1 and less than about 2 vCPU when the addons are on, or document a minimum pool size.

Lingua principale
Go
Stelle
2.2k
Fork
209
Merge medio
3g 10h
PR unite (30g)
237

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di cozystack/cozystack

Tutte le issue di cozystack/cozystack

Issue simili

Altre issue su Go

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.