[backport/2.3] Remove distutils from connection plugin (#456)
Remove distutils from connection plugin
SUMMARY
distutils.spawn.find_executable is deprecated and shutils.which is a
suitable replacement.
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: Gonéri Le Bouder goneri@lebouder.net
Reviewed-by: Joseph Torcasso
(cherry picked from commit 531a9fe)
[backport/2.3] k8s_cp - fix issue when using local_path (#422)
Depends-On: #446
k8s_cp - fix issue when using local_path
SUMMARY
When copying from local path to pod, the file is found on the controller node instead of the managed node.
This PR aims to resolve this issue.
Fixes#421
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
k8s_cp
Reviewed-by: Abhijeet Kasurde
Reviewed-by: Mike Graves mgraves@redhat.com
(cherry picked from commit 1d05cf5)
[backport/2.3] k8s - fix issue when try to delete resources using label_selectors op…
Depends-On: #446
k8s - fix issue when try to delete resources using label_selectors
SUMMARY
The kubernetes dynamic client has label_selector parameter for the delete method, however based on the documentation of REST API we cannot delete resources using labelSelector option, this fix update the way the resources are deleted. The list of resources are deleted one after another like in the kubectl go client.
Fixes#428
ISSUE TYPE
Bugfix Pull Request
Reviewed-by: Abhijeet Kasurde
(cherry picked from commit f2f4b66)
[backport/2.3] continue waiting when an exception is raised (#408)
Depends-On: #446
Continue waiting when an exception is raised
SUMMARY
When an exception is raised and the wait_timeout is not reached, we should continue waiting as this may occurs due to temporary issue on cluster
Fixes#407
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: Mike Graves mgraves@redhat.com
Reviewed-by: Abhijeet Kasurde
(cherry picked from commit f418353)
[backport/2.3] Change line in doc fragment yaml (#439)
Depends-On: #446
Change line in doc fragment yaml
SUMMARY
For whatever reason, the one line in this doc fragment leads to sanity
failures in the redhat.openshift collection, which uses this fragment.
The downstream build process for that collection creates yaml that
appears to be valid, but that fails to lint. I'm not sure exactly which
tool the problem is in, but the easiest solution is to just remove the
single quotes here.
ISSUE TYPE
Docs Pull Request
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: Abhijeet Kasurde
(cherry picked from commit b5cfc85)
[backport/2.3] Remove omit from template resource (#432)
Depends-On: #446
Remove omit value from template args
SUMMARY
While defining resource using template parameter, the code does not remove the omit value if any.
This fix adds a post process to remove any omit value from the resource definition.
fixes#431
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
k8s*
Reviewed-by: Mike Graves mgraves@redhat.com
Reviewed-by: Abhijeet Kasurde
(cherry picked from commit 882e672)
[backport/2.3] fix issue when using k8s_drain with disable_eviction set to yes (#418)
Depends-On: #446
fix issue when using k8s_drain with disable_eviction set to yes
SUMMARY
fixes#416
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
k8s_drain
ADDITIONAL INFORMATION
Reviewed-by: Abhijeet Kasurde
(cherry picked from commit 074f0a6)
[backport/2.3] Upgrade black version (#424)
Depends-On: ansible/ansible-zuul-jobs#1515
Upgrade black version
SUMMARY
Move off of beta version of black and pin to current calendar year
version.
The only manual changes here are to tox.ini. Everything else is from running the new version of black.
ISSUE TYPE
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: Abhijeet Kasurde
(cherry picked from commit 7c71436)
Fix waiting on StatefulSet scale down
SUMMARY
When scaling a StatefulSet down to 0 replicas the wait will fail
because some properties of the status (readyReplicas, updatedReplicas)
will not exist. These are probably defined as omitempty in the API and
since the value is zero are not present in the response.
Fixes#203
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
k8s_scale
ADDITIONAL INFORMATION
Reviewed-by: Gonéri Le Bouder <goneri@lebouder.net>
Fix validation errors in plugin documentation
Depends-On: ansible/ansible-zuul-jobs#1385
SUMMARY
This fixes validation errors in plugin documentation now that ansible
test for 2.13 is running validate-modules on all plugins. The kubectl
connection plugin validation is ignored because there seems to be a
requirement for the author field to have a github username, which we do
not have.
ISSUE TYPE
Docs Pull Request
COMPONENT NAME
plugins/connection/kubectl
plugins/inventory/k8s
plugins/lookup/k8s
plugins/lookup/kustomize
ADDITIONAL INFORMATION
Reviewed-by: None <None>
helm_template: add optional show_only and release_namespace arguments
SUMMARY
This PR adds the "show_only" and "release_namespace" as optional arguments to the helm_template module.
It does some work towards #313.
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
changelogs/fragments/313-helm-template-add-support-for-show-only-and-release-namespace.yml
plugins/modules/helm_template.py
tests/unit/modules/test_helm_template.py
ADDITIONAL INFORMATION
The PR does include unit tests instead of integration test.
Reasoning:
The existing integration tests already include a task based on helm_template. So we know that the module does a proper job of using the command line generated inside the module to call helm.
As I trust helm itself to "do its job" correctly, all that should be necessary is to test the correct generation of the command line itself. The included unit tests hopefully do a proper job.
With regards of the pretty long testing times for the module, I really prefer unit tests, if at all possible.
Please let me know if this fits.
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
k8s_scale, k8s_rollback - add support for check_mode
SUMMARY
closes#243 and #244
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
k8s_scale
k8s_rollback
ADDITIONAL INFORMATION
Reviewed-by: Alina Buzachis <None>
Reviewed-by: None <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Adding sleep 0 as workaround when copying files with kubectl exec
SUMMARY
For all the commands executed remotely, ** && sleep 0** will be
appended as a workaround for all the commands to terminate properly:
16def8050a/lib/ansible/plugins/action/__init__.py (L1243)
Workaround will be applied in case of kubectl exec too:
kubernetes.core/plugins/connection/kubectl.py
Line 300
in
b19ff9d
super(Connection, self).exec_command(cmd, in_data=in_data, sudoable=sudoable)
That is not the case in the case of the file copy executed by using kubectl exec, therefore it is possible for the kubectl exec to
terminate before dd finishes properly causing the file to be truncated.
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
changelogs/fragments/321-kubectl_sleep.yml
plugins/connection/kubectl.py
k8s_exec: Select first container from the pod
SUMMARY
kubectl command select first container from the pod in order
to execute commands on. We replicate the same behavior in k8s_exec
module.
Fixes: #358
Signed-off-by: Abhijeet Kasurde akasurde@redhat.com
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
changelogs/fragments/358-k8s_exec.yml
plugins/modules/k8s_exec.py
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
Use resource prefix when apiVersion is v1
SUMMARY
When getting a resource from the core api group, the prefix was not
passed, leading the lookup to happen in all api groups. This broad
search is not really necessary and leads to problems in some corner
cases, for example, when an api is deleted after the api group list is
cached.
This fix uses the 'api' prefix when the apiVersion is 'v1', as this is
almost certainly what the user wants. As a fallback, to retain backwards
compatibility, the old behavior is used if the first lookup failed to
find a resource. Given that the module defaults to 'v1' for the
apiVersion, there are likely many cases where a resource, such as
StatefulSet, is used while failing to provide an apiVersion. While
technically incorrect, this has worked in most cases, so we probably
shouldn't break this behavior.
Fixes#351
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
changelogs/fragments/364-use-resource-prefix.yaml
plugins/module_utils/common.py
helm_repository: Added support for common options
SUMMARY
Added support for host, api_key, ca_cert,
and validate_certs in helm_repository module.
Signed-off-by: Abhijeet Kasurde akasurde@redhat.com
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
changelogs/fragments/helm_repository.yml
plugins/modules/helm_repository.py
helm_template: change order of values_release and values_files
SUMMARY
This fix aligns precedence of release_values and values_files in kubernetes.core.helm_template with the one in kubernetes.core.helm.
The values in release_values are now processed last, thus with the highest precedence.
This allows overwriting of values in values_files with the values presented in release_values.
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
changelogs/fragments/348-helm_template-fix-precedence-of-release-values-over-values-files.yaml
plugins/modules/helm_template.py
tests/unit/modules/test_helm_template.py
Prepare for distutils.version being removed in Python 3.12
SUMMARY
distutils has been deprecafed and will be removed from
Python's stdlib in Python 3.12 (see python.org/dev/peps/pep-0632).
This PR replaces the use of distutils.version.LooseVersion and distutils.version.StrictVersion
with LooseVersion from the vendored copy of distutils.version
included with ansible-core 2.12 (ansible/ansible#74644) if available,
and falls back to distutils.version for ansible-core 2.11 and before.
Since ansible-core 2.11 and earlier do not support Python 3.12 (since
they use LooseVersion itself in various places), this incomplete fix
should be OK for now. Also, the way this PR works (by adding a new
module_utils version that abstracts away where LooseVersion comes from),
it is easy to also fix this for ansible-core 2.11 and earlier later on.
Signed-off-by: Abhijeet Kasurde akasurde@redhat.com
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
changelogs/fragments/disutils.version.yml
molecule/default/roles/helm/library/helm_test_version.py
plugins/module_utils/common.py
plugins/module_utils/version.py
plugins/modules/helm.py
Reviewed-by: Felix Fontein <felix@fontein.de>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
Documentation update for kubernetes.core.helm
Clarify usage of the module for doing helm repo update only.
I used collection_prep_add_docs as explained in CONTRIBUTING.md, not sure if
that's correct ?
Fixes#316
@Akasurde
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
Add delete_emptydir_data to drain delete_options
SUMMARY
Adds delete_emptydir_data option to k8s_drain.delete_options to evict pods with an emptyDir volume attached.
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
k8s_drain
ADDITIONAL INFORMATION
Be gentle, this is my first pull request 😨
Basically adds the kubectl drain <node> --delete-emptydir-data feature, including tests.
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: Jorn Eilander <None>
Reviewed-by: None <None>
Reviewed-by: None <None>
add support for community.okd.openshift_adm_groups_sync
SUMMARY
new module community.okd.openshift_adm_group_sync requires action group
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
k8s - add support for Server Side apply
SUMMARY
Server side apply is now support for k8s module with this Pull request.
The feature is not yet released on kubernetes-client, once this is done, we can merge this pull request.
closes#87
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
k8s
ADDITIONAL INFORMATION
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
Reviewed-by: None <None>
Fix for common non-ASCII characters in CRDs
This should keep the module safe from digesting non-ASCII chars like here (https://github.com/projectcalico/api/pull/46/files)
SUMMARY
Add support for non-ASCII chars in manifests.
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
core.k8s module failing if resources contain non ascii chars
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: Alessandro Rossi <None>
Reviewed-by: None <None>
Return diff in helm check mode
When the helm module is executed in check mode with the helm diff plugin
installed, it now returns the diff.
SUMMARY
When the helm module is executed in check mode with the helm diff plugin
installed, it now returns the diff.
COMPONENT NAME
helm
ADDITIONAL INFORMATION
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: Fabrice <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
Fix k8s_drain failing when pod has local storage
SUMMARY
The module fails to define the pod_names variable before using it for
pods with local storage.
Fixes#292
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
k8s_drain
ADDITIONAL INFORMATION
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: None <None>
Reviewed-by: None <None>
helm: add pass-credentials key
SUMMARY
In helm version v3.6.1 when downloading charts from password protected repositories that served from a different domain than the repository, need to use --pass-credentials key.
Add possibility to use the pass-credentials key in helm_repository.py
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
helm_repository
Reviewed-by: None <None>
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
add support for user impersonation for k8s modules
SUMMARY
k8s module should not allow user to perform operation using impersonation as describe here
https://kubernetes.io/docs/reference/access-authn-authz/authentication/#user-impersonation
This pull request closes#40
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: Abhijeet Kasurde <None>
Reviewed-by: None <None>
add no_proxy support to k8s*
SUMMARY
close#271
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
plugins/module_utils/args_common.py
plugins/modules/k8s*
ADDITIONAL INFORMATION
It requires latest kubernetes library(>=19.15.0) to use this feature.
pip install kubernetes>=19.15.0
then, use following snippet yaml:
- k8s:
state: present
src: "deployment.yaml"
proxy: "http://proxy.yourdomain.com:8080/"
no_proxy: "localhost,.yourdomain.com,127.0.0.0/8,10.0.0.0/8,172.16.0.0/12,192,168.0.0/16"
or use environment variable K8S_AUTH_NO_PROXY as well as K8S_AUTH_PROXY.
Reviewed-by: None <None>
Reviewed-by: None <None>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
Don't wait on *List resources for info module
SUMMARY
We can't use the same wait logic on *List resources because they lack
the same metadata that other resources have. We should ensure that we
are waiting on the items in the list, but not the list itself. Waiting
on the list itself results in unexpected behavior.
This fixes the waiting logic when waiting on a list to wait until the
list being queried contains one or more items, or the wait timeout has
been reached. Each item in the list can then be waited on with the usual
wait logic.
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: None <None>
Reviewed-by: Alina Buzachis <None>
Reviewed-by: None <None>
Add support for dry run
SUMMARY
Kubernetes server-side dry run will be used when the kubernetes client
version is >=18.20.0. For older versions of the client, the existing
client side speculative change implementation will be used.
The effect of this change should be mostly transparent to the end user
and is reflected in the fact the tests have not changed but should still
pass. With this change, there are a few edge cases that will be
improved. One example of these edge cases is to use check mode on an
existing Service resource. With dry run this will correctly report no
changes, while the older client side implementation will erroneously
report changes to the port spec.
ISSUE TYPE
Feature Pull Request
COMPONENT NAME
ADDITIONAL INFORMATION
Reviewed-by: Gonéri Le Bouder <goneri@lebouder.net>
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: Alina Buzachis <None>
Reviewed-by: None <None>
Reviewed-by: None <None>
common/_wait_for: ensure label_selectors is optional
Depends-On: ansible/ansible-zuul-jobs#1125
The label_selectors is a new parameter for _wait_for that was
introduced in #158.
The value is new and it can be set to None to make it optional. It should
not be mandatory a non optional parameter.
Reviewed-by: None <None>
Reviewed-by: Alina Buzachis <None>
Reviewed-by: None <None>
fix k8s_cp uploading when deployed container's WORKDIR is other than '/'
SUMMARY
fix#222
ISSUE TYPE
Bugfix Pull Request
COMPONENT NAME
k8s_cp
ADDITIONAL INFORMATION
Reviewed-by: Mike Graves <mgraves@redhat.com>
Reviewed-by: None <None>
Reviewed-by: None <None>
