The fallback_ips tasks are essentially serializing the gathering of one
fact on all the hosts, which can have dramatic performance implications
on large clusters (several minutes).
This is essentially a reversal of 35f248dff0
Being able to run without refreshing the cache facts is not worth it.
We keep fallback_ip for now, simply changing the access to a normal
hostvars variable instead of a custom dictionnary.
* Update cluster-role for cilium to prevent errors in agent startup
ciliumloadbalancerippools permissions exists in the cilium helm chart for version 1.13.0
https://github.com/cilium/cilium/blob/v1.13.0/install/kubernetes/cilium/templates/cilium-agent/clusterrole.yaml#L71
The agent also needs permissions to read/watch secrets for bgp auth secrets when using CiliumBGPPeeringPolicy with a secret.
* Remove list/watch permissions for secrets
* Remove secrets from list/watch permissions
The old repository for these has been deleted, leaving the previous
configuration not possible to deploy, and even currently running clusters
fail after a restart as the DeameonSet has ImagePullPolicy: Always. More
details can be found here: kubernetes-sigs/vsphere-csi-driver#3053
As of writing, only CSI driver versions 3.1.2 to 3.3.1 is available in
this registry. This "officially" supports Kubernetes 1.26 to 1.30. Since
older drivers are not available, I have removed some feature-gating for
those unavailable versions while I was at it. For the cloud provider,
the `latest` image is now missing, and only 1.28.0 to 1.31.0 are
available. I've set the latest of these as the new default.
I also updated the documented default versions, as they were all out of
date and not aligned with actual code defaults.
Nodes to api-server relies by default certificates, and bootstrap
tokens, and there should be no need to generate tokens for every nodes,
even when enabling static token auth.
Testing for group membership with group names makes Kubespray more
tolerant towards the structure of the inventory.
Where 'inventory_hostname in groups["some_group"] would fail if
"some_group" is not defined, '"some_group" in group_names' would not.
Specifying one directory for kubeadm patches is not ideal:
1. It does not allow working with multiples inventories easily
2. No ansible templating of the patch
3. Ansible path searching can sometimes be confusing
Instead, provide the patch directly in a variable, and add some quality
of life to handle components targeting and patch ordering more
explicitly (`target` and `type` which are translated to the kubeadm
scheme which is based on the file name)
kubernetes/control-plane and kubernetes/kubeadm roles both push kubeadm
patches in the same way.
Extract that code and make it a dependency of both.
This is safe because it's only configuration for kubeadm, which only
takes effect when kubeadm is run.
* Update multus to v4.1.0 and clarify cilium compatibility
* Fix: bug introduced by #10934 where the template would break if multus was defined
* Set priorityClassName to system-node-critical for multus pods
Remove system|kube_master_<resource>_reserved variables.
Those variables are unnecessary because users can simply use the
variables in group_vars if they which to differentiate control plane
nodes from other nodes.
Set conservative defaults for ephemeral-storage and pids for both kube
and system reserved resources.
Working symlinks are dependant on git configuration (when using the playbook as
a git repository, which is common), precisely `git config
core.symlinks`.
While this is enabled by default, some company policies will disable it.
Instead, use import_tasks which should avoid that class of bugs.
* Simplify docker systemd unit
systemd handles missing unit by ignoring the dependency so we don't need
to template them.
* Remove RHEL 7/CentOS 7 support
- remove ref in kubespray roles
- move CI from centos 7 to 8
- remove docs related to centos7
* Remove container-storage-setup
Only used for RHEL 7 and CentOS 7
* Feat: add Gateway API CRDs installation
Signed-off-by: ChengHao Yang <17496418+tico88612@users.noreply.github.com>
* Feat: add Gateway API CRDs variable in inventory
Signed-off-by: ChengHao Yang <17496418+tico88612@users.noreply.github.com>
---------
Signed-off-by: ChengHao Yang <17496418+tico88612@users.noreply.github.com>
* make calico api server manifest backward compatible with version older than 3.27.3
Add 3.28.1 checksums
Add 3.28.0 checksums
Change default version to 3.27.3
* change default calico version to 3.28.1
* Set mount type to DirectoryOrCreate for hostPath needed by Calico
Fixes https://github.com/kubernetes-sigs/kubespray/issues/10947
This patch aims to be minimal and intentionally:
- does not change the generation logic for `supersede_domain` and `supersede_search`
- does not change how `nameserverentries` (for NetworkManager) is built
It seems like `nameserverentries` in the "Generate nameservers for resolvconf, including cluster DNS"
task is built the same way as `dhclient_supersede_nameserver_entries_list`.
However, `nameserverentries` in the "Generate nameservers for resolvconf, not including cluster DNS"
task (below) is built differently for some reason. It includes `configured_nameservers` as well.
Due to these differences, I have refrained from reusing the same building logic
(`dhclient_supersede_nameserver_entries_list`) for both.
If the `configured_nameservers` addition can be removed or made to apply
to dhclient as well, we could potentially build a single list and then
generate the `nameserverentries` and `supersede_nameserver` strings from it.
ChengHao Yang
Max Gautier
Kubernetes Prow Robot
Baargav
Olivier Boudry
peterw
kyrie
Kay Yan
Serge Hartmann
Erwan Miran
M. Hamzah Khan
DirkTheDaring
janosbabik
Bakke
Philip Sabri
Leeon Fu
Qasim Mehmood
Kevin Huang
Bogdan Sass
刘旭
Jongwoo Han
ERIK
Lihai Tu
Ehsan Golpayegani
Takuya Murakami
Selçuk Arıbalı
Vlad Korolev
Lola Delannoy
Mohamed Omar Zaian
kokyhm
Ho Kim
Mengxin Liu
Cyclinder
Not Darko
Slavi Pantaleev
Injun Baeg
James
Sanyam Shah
tu1h
R. P. Taylor
Ugur Can Ozturk
Tom M.
Bas