Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

change nodeSelector label from deprecated beta.kubernetes.io/os and arch to kubernetes.io prefix (#7138)

TASK [Generate a list of information about the images on a node]
registers list of container images to docker_images.
Then the next TASK [Set pull_required if the desired image is not
yet loaded] does based on expecting images are registered.
However sometimes the first TASK was failed as [1] but the failure
is ignored due to failed_when:false and it makes another issue.
This removes this unnecessary failed_when to detect the failure
at the point.
In addition, this removes no_log:true also because the output doesn't
contain any sensitive data and now it just makes debugging difficult.

[1]: https://gitlab.com/kargo-ci/kubernetes-sigs-kubespray/-/jobs/934714534#L2953

no_proxy is a pain to get right, and having proxy variables present causes issues
(k8s components get proxy configuration after upgrade, see #7100)

It's better to only configure what require proxy:
- the runtime (containerd/docker/crio)
- the package manager + apt_key
- the download tasks

Tested with the following clusters
- 4 CentOS 8 nodes
- 1 Ubuntu 20.04 node

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

In some environments, it might not be possible to ping the IP address
of the nodes, e.g., because ICMP echo is blocked.

This commit allows kubespray to be configured to disable the ping
check, while performing all other checks.

TASK [network_plugin/calico : Calico | Configure calico network pool] **********
task path: /builds/kargo-ci/kubernetes-sigs-kubespray/roles/network_plugin/calico/tasks/install.yml:138
Friday 08 January 2021  17:10:12 +0000 (0:00:01.521)       0:11:36.885 ********
[WARNING]: The value {'kind': 'IPPool', 'apiVersion': 'projectcalico.org/v3',
'metadata': {'name': 'default-pool'}, 'spec': {'blockSize': 24, 'cidr':
'10.233.64.0/18', 'ipipMode': 'Always', 'vxlanMode': 'Never', 'natOutgoing':
True}} (type dict) in a string field was converted to "{'kind': 'IPPool',
'apiVersion': 'projectcalico.org/v3', 'metadata': {'name': 'default-pool'},
'spec': {'blockSize': 24, 'cidr': '10.233.64.0/18', 'ipipMode': 'Always',
'vxlanMode': 'Never', 'natOutgoing': True}}" (type string). If this does not
look like what you expect, quote the entire value to ensure it does not change.

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

Signed-off-by: Brendan Holmes <5072156+holmesb@users.noreply.github.com>

* Improve how we set 'proxy=' in yum.conf or dnf.conf

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

* Fixup spaces in no_proxy

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

* Add svc,svc.{{ dns_domain }} to no_proxy

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

Valuating conditional (need_https_proxy.rc != 0) fail if http_proxy set and skip_http_proxy_on_os_packages is true (#7078)

* Remove because of empty need_http_proxy.rc if http/https_proxy and skip_http_proxy_on_os_packages=true is set

* Modify sample for debian and centos skip_http_proxy

* Modify sample for debian and centos skip_http_proxy

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

This fixes 1945499e

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

All packages have proper dependencies in latest versions

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

If some settings were changed from the default but not commited into an inventory repo,
we risk breaking the cluster / cause downtime, so add some extra checks

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

This fixes markdownlint failures under roles/

Upgrading docker / containerd without adapting the configuration might break the node,
so disable docker-ce repo by default.
We are already using dpkg hold for Debian.

All containerd.io packages provide /usr/bin/runc, so no need to check

yum_conf was never used for containerd

module_hotfixes should not be needed with the EL8 repo

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

See https://github.com/projectcalico/calico/issues/3271

Otherwise Calico can get into a fight with NM about who "owns" the vxlan.calico
interface, breaking all pod traffic.

* Fedora CoreOS: Fix for ethtool pre-installed

Fix error in rpm-ostree when ethtool is already insatlled (FCOS >= 32.20201104.3.0)

* Fedora CoreOS: Fix connection lost

Fedora CoreOS: Ignore connection lost due to reboot and continues the playbook

* download run once feature for CRI-O

* fix typo

* fix test

We are currently setting the IP variable to hostIP,
Before https://github.com/projectcalico/node/pull/593

 (not yet released)
Calico interpret that as hostIP/32
Using 'can-reach' we get the future behavior
This fixes vxlan and IPIP CrossSubnet modes

Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>

Signed-off-by: François Travais <francois@travais.fr>

* SHAs for 1.19.6, 1.18.14, 1.18.13, 1.17.16 and 1.17.15

* Fix etcd version in README

* Bump kube_version to 1.19.6

* update files to handle multi-asn bgp peering conditions.

* put back in the serviceClusterIPs.  Bad merge.

* remove extraneous environment var.

* update files as discussed with mirwan

* update titles.

* add not in.

* add a conditional for using bgp to advertise cluster ips.

Co-authored-by: marlow-h <mweston@habana.ai>

If cluster-name is not set, the default value "kubernetes" is used.
The loadbalancees created by Kubernetes follow the format:
  kube_service_clusterName_serviceNamespace_serviceName
If 2 clusters create a loadbalancer for the same service in the same
namespace, they will share the same non-working loadbalancer.

Signed-off-by: Cedric Hnyda <cedric.hnyda@itera.io>

When opening the main.yaml, vi cannot show the string correctly
due to non-ascii string. This replaces it.

* Update hashes and set default version to 1.19.5

Signed-off-by: anthr76 <hello@anthonyrabbito.com>

* Reorder hashes

1.19.5 hashes should be near 1.19.x

* Added back blank line

Since ansible 2.9 search cannot be used as filter after a pipe but after `is`

Signed-off-by: Sylvain Desbureaux <sylvain.desbureaux@orange.com>

This fixes the following warning:

  [kubernetes/client : Generate admin kubeconfig with external api endpoint]
  [WARNING]: Consider using the file module with state=directory rather than
  running 'mkdir'.  If you need to use command because file is insufficient
  you can

* add support crio version for various k8s vers

* regexp in pkg versions