search

kubernetes-sigs / kubespray

Deploy a Production Ready Kubernetes Cluster
Apache License 2.0
15.86k stars 6.41k forks source link

download role executed too early when CRI-O instructions are followed #10658

Open dkasanic opened 9 months ago

dkasanic commented 9 months ago

Environment:

Kubespray version (commit) (git rev-parse --short HEAD): 3acacc615

Network plugin used: calico

Full inventory with variables (ansible -i inventory/sample/inventory.ini all -m debug -a "var=hostvars[inventory_hostname]"):

skip_downloads: false

Output of ansible run:

2023-11-23 00:44:05,460 p=2875882 u=root n=ansible | TASK [kubernetes_sigs.kubespray.download : Prep_kubeadm_images | Create kubeadm config] ***
2023-11-23 00:44:05,461 p=2875882 u=root n=ansible | fatal: [aj09-17-dell-spr]: FAILED! => {
    "changed": false,
    "checksum": "601e28489b672da953a83dc549261b385c01a692"
}

MSG:

Destination directory /etc/kubernetes does not exist

Anything else do we need to know: Download role got called as dependency from kubespray-defaults role but skip_downloads: true var defined in meta/main.yml was not applied. That results in downloading items early on when /etc/kubernetes directory does not exist yet,

yankay commented 9 months ago

HI @dkasanic

Thanks for the issue and PR. Would you please give more information about the kubespray or ansible config to reproduce the error ?

It's very helpful :-)

Thanks you :-)

dkasanic commented 9 months ago

Hello, @yankay

In my env, I install kubespray as galaxy collection and then import cluster.yml. To reproduce the error, I believe following snippet of tasks in my playbook is enough:

- name: add crio runtime vars
   set_fact:
     container_manager: crio
     download_container: false
     skip_downloads: false
     etcd_deployment_type: host
- name: Deploy cluster via Kubespray
  any_errors_fatal: true
  ansible.builtin.import_playbook: kubernetes_sigs.kubespray.cluster

It seems in such case, skip_downloads: true var definition in meta/main.yml file will not kick in properly and the download role will start download items, but that should not happen at this stage of run. It should happen after kubespray-defaults role is executed and download role is called from cluster.yaml playbook.

As soon as I removed skip_downloads: false var definition from set_fact task, deployment started working correctly. The problem is in following meta/main.yml file:

dependencies:
  - role: download
    skip_downloads: true
    tags:
      - facts

as per ansible docs, it should be defined as:

dependencies:
  - role: download
    vars:
      skip_downloads: true
    tags:
      - facts
VannTen commented 9 months ago

Does #10626 fix your problem ? (since download is no longer pulled in by kubespray-defaults)

VannTen commented 7 months ago

Is the problem still present on master ? I believe the PR linked in the previous message might have fixed the issue

Download role got called as dependency from kubespray-defaults role but skip_downloads: true var defined in meta/main.yml was not applied. That results in downloading items early on when /etc/kubernetes directory does not exist yet,

(Since this is no longer true)

VannTen commented 7 months ago

/triage needs-information

k8s-triage-robot commented 4 months ago

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

You can:

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

k8s-triage-robot commented 3 months ago

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

You can:

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten

vaibhav2107 commented 3 months ago

/remove-lifecycle rotten

k8s-triage-robot commented 18 hours ago

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

You can:

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale