Files
community.okd/plugins/module_utils/openshift_adm_prune_images.py
Bikouo Aubin a63e5b7b36 Update CI - Continue work from #195 (#202)
* Upgrade Ansible and OKD versions for CI

* Use ubi9 and fix sanity

* Use correct pip install

* Try using quotes

* Ensure python3.9

* Upgrade ansible and molecule versions

* Remove DeploymentConfig

DeploymentConfigs are deprecated and seem to now be causing idempotence
problems. Replacing them with Deployments fixes it.

* Attempt to fix ldap integration tests

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Move sanity and unit tests to GH actions

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Firt round of sanity fixes

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Add kubernetes.core collection as sanity requirement

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Add ignore-2.16.txt

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Attempt to fix units

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Add ignore-2.17

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Attempt to fix unit tests

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Add pytest-ansible to test-requirements.txt

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Add changelog fragment

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Add workflow for ansible-lint

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Apply black

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Fix linters

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Add # fmt: skip

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Yet another round of linting

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Yet another round of linting

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Remove setup.cfg

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Revert #fmt

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Use ansible-core 2.14

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Cleanup ansible-lint ignores

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>

* Try using service instead of pod IP

* Fix typo

* Actually use the correct port

* See if NetworkPolicy is preventing connection

* using Pod internal IP

* fix adm prune auth roles syntax

* adding some retry steps

* fix: openshift_builds target

* add flag --force-with-deps when building downstream collection

* Remove yamllint from tox linters, bump minimum python supported version to 3.9, Remove support for ansible-core < 2.14

---------

Signed-off-by: Alina Buzachis <abuzachis@redhat.com>
Co-authored-by: Mike Graves <mgraves@redhat.com>
Co-authored-by: Alina Buzachis <abuzachis@redhat.com>
2023-11-15 17:00:38 +00:00

496 lines
19 KiB
Python

#!/usr/bin/env python
from __future__ import absolute_import, division, print_function
__metaclass__ = type
from datetime import datetime, timezone, timedelta
import copy
from ansible.module_utils._text import to_native
from ansible.module_utils.parsing.convert_bool import boolean
from ansible.module_utils.six import iteritems
from ansible_collections.community.okd.plugins.module_utils.openshift_common import (
AnsibleOpenshiftModule,
)
from ansible_collections.community.okd.plugins.module_utils.openshift_images_common import (
OpenShiftAnalyzeImageStream,
get_image_blobs,
is_too_young_object,
is_created_after,
)
from ansible_collections.community.okd.plugins.module_utils.openshift_docker_image import (
parse_docker_image_ref,
convert_storage_to_bytes,
)
try:
from kubernetes import client
from kubernetes.client import rest
from kubernetes.dynamic.exceptions import (
DynamicApiError,
NotFoundError,
ApiException,
)
except ImportError:
pass
ApiConfiguration = {
"LimitRange": "v1",
"Pod": "v1",
"ReplicationController": "v1",
"DaemonSet": "apps/v1",
"Deployment": "apps/v1",
"ReplicaSet": "apps/v1",
"StatefulSet": "apps/v1",
"Job": "batch/v1",
"CronJob": "batch/v1beta1",
"DeploymentConfig": "apps.openshift.io/v1",
"BuildConfig": "build.openshift.io/v1",
"Build": "build.openshift.io/v1",
"Image": "image.openshift.io/v1",
"ImageStream": "image.openshift.io/v1",
}
def read_object_annotation(obj, name):
return obj["metadata"]["annotations"].get(name)
def determine_host_registry(module, images, image_streams):
# filter managed images
def _f_managed_images(obj):
value = read_object_annotation(obj, "openshift.io/image.managed")
return boolean(value) if value is not None else False
managed_images = list(filter(_f_managed_images, images))
# Be sure to pick up the newest managed image which should have an up to date information
sorted_images = sorted(
managed_images, key=lambda x: x["metadata"]["creationTimestamp"], reverse=True
)
docker_image_ref = ""
if len(sorted_images) > 0:
docker_image_ref = sorted_images[0].get("dockerImageReference", "")
else:
# 2nd try to get the pull spec from any image stream
# Sorting by creation timestamp may not get us up to date info. Modification time would be much
sorted_image_streams = sorted(
image_streams,
key=lambda x: x["metadata"]["creationTimestamp"],
reverse=True,
)
for i_stream in sorted_image_streams:
docker_image_ref = i_stream["status"].get("dockerImageRepository", "")
if len(docker_image_ref) > 0:
break
if len(docker_image_ref) == 0:
module.exit_json(changed=False, result="no managed image found")
result, error = parse_docker_image_ref(docker_image_ref, module)
return result["hostname"]
class OpenShiftAdmPruneImages(AnsibleOpenshiftModule):
def __init__(self, **kwargs):
super(OpenShiftAdmPruneImages, self).__init__(**kwargs)
self.max_creation_timestamp = self.get_max_creation_timestamp()
self._rest_client = None
self.registryhost = self.params.get("registry_url")
self.changed = False
def list_objects(self):
result = {}
for kind, version in iteritems(ApiConfiguration):
namespace = None
if self.params.get("namespace") and kind.lower() == "imagestream":
namespace = self.params.get("namespace")
try:
result[kind] = self.kubernetes_facts(
kind=kind, api_version=version, namespace=namespace
).get("resources")
except DynamicApiError as e:
self.fail_json(
msg="An error occurred while trying to list objects.",
reason=e.reason,
status=e.status,
)
except Exception as e:
self.fail_json(
msg="An error occurred while trying to list objects.",
error=to_native(e),
)
return result
def get_max_creation_timestamp(self):
result = None
if self.params.get("keep_younger_than"):
dt_now = datetime.now(timezone.utc).replace(tzinfo=None)
result = dt_now - timedelta(minutes=self.params.get("keep_younger_than"))
return result
@property
def rest_client(self):
if not self._rest_client:
configuration = copy.deepcopy(self.client.configuration)
validate_certs = self.params.get("registry_validate_certs")
ssl_ca_cert = self.params.get("registry_ca_cert")
if validate_certs is not None:
configuration.verify_ssl = validate_certs
if ssl_ca_cert is not None:
configuration.ssl_ca_cert = ssl_ca_cert
self._rest_client = rest.RESTClientObject(configuration)
return self._rest_client
def delete_from_registry(self, url):
try:
response = self.rest_client.DELETE(
url=url, headers=self.client.configuration.api_key
)
if response.status == 404:
# Unable to delete layer
return None
# non-2xx/3xx response doesn't cause an error
if response.status < 200 or response.status >= 400:
return None
if response.status != 202 and response.status != 204:
self.fail_json(
msg="Delete URL {0}: Unexpected status code in response: {1}".format(
response.status, url
),
reason=response.reason,
)
return None
except ApiException as e:
if e.status != 404:
self.fail_json(
msg="Failed to delete URL: %s" % url,
reason=e.reason,
status=e.status,
)
except Exception as e:
self.fail_json(msg="Delete URL {0}: {1}".format(url, type(e)))
def delete_layers_links(self, path, layers):
for layer in layers:
url = "%s/v2/%s/blobs/%s" % (self.registryhost, path, layer)
self.changed = True
if not self.check_mode:
self.delete_from_registry(url=url)
def delete_manifests(self, path, digests):
for digest in digests:
url = "%s/v2/%s/manifests/%s" % (self.registryhost, path, digest)
self.changed = True
if not self.check_mode:
self.delete_from_registry(url=url)
def delete_blobs(self, blobs):
for blob in blobs:
self.changed = True
url = "%s/admin/blobs/%s" % (self.registryhost, blob)
if not self.check_mode:
self.delete_from_registry(url=url)
def update_image_stream_status(self, definition):
kind = definition["kind"]
api_version = definition["apiVersion"]
namespace = definition["metadata"]["namespace"]
name = definition["metadata"]["name"]
self.changed = True
result = definition
if not self.check_mode:
try:
result = self.request(
"PUT",
"/apis/{api_version}/namespaces/{namespace}/imagestreams/{name}/status".format(
api_version=api_version, namespace=namespace, name=name
),
body=definition,
content_type="application/json",
).to_dict()
except DynamicApiError as exc:
msg = "Failed to patch object: kind={0} {1}/{2}".format(
kind, namespace, name
)
self.fail_json(msg=msg, status=exc.status, reason=exc.reason)
except Exception as exc:
msg = "Failed to patch object kind={0} {1}/{2} due to: {3}".format(
kind, namespace, name, exc
)
self.fail_json(msg=msg, error=to_native(exc))
return result
def delete_image(self, image):
kind = "Image"
api_version = "image.openshift.io/v1"
resource = self.find_resource(kind=kind, api_version=api_version)
name = image["metadata"]["name"]
self.changed = True
if not self.check_mode:
try:
delete_options = client.V1DeleteOptions(grace_period_seconds=0)
return resource.delete(name=name, body=delete_options).to_dict()
except NotFoundError:
pass
except DynamicApiError as exc:
self.fail_json(
msg="Failed to delete object %s/%s due to: %s"
% (kind, name, exc.body),
reason=exc.reason,
status=exc.status,
)
else:
existing = resource.get(name=name)
if existing:
existing = existing.to_dict()
return existing
def exceeds_limits(self, namespace, image):
if namespace not in self.limit_range:
return False
docker_image_metadata = image.get("dockerImageMetadata")
if not docker_image_metadata:
return False
docker_image_size = docker_image_metadata["Size"]
for limit in self.limit_range.get(namespace):
for item in limit["spec"]["limits"]:
if item["type"] != "openshift.io/Image":
continue
limit_max = item["max"]
if not limit_max:
continue
storage = limit_max["storage"]
if not storage:
continue
if convert_storage_to_bytes(storage) < docker_image_size:
# image size is larger than the permitted limit range max size
return True
return False
def prune_image_stream_tag(self, stream, tag_event_list):
manifests_to_delete, images_to_delete = [], []
filtered_items = []
tag_event_items = tag_event_list["items"] or []
prune_over_size_limit = self.params.get("prune_over_size_limit")
stream_namespace = stream["metadata"]["namespace"]
stream_name = stream["metadata"]["name"]
for idx, item in enumerate(tag_event_items):
if is_created_after(item["created"], self.max_creation_timestamp):
filtered_items.append(item)
continue
if idx == 0:
istag = "%s/%s:%s" % (
stream_namespace,
stream_name,
tag_event_list["tag"],
)
if istag in self.used_tags:
# keeping because tag is used
filtered_items.append(item)
continue
if item["image"] not in self.image_mapping:
# There are few options why the image may not be found:
# 1. the image is deleted manually and this record is no longer valid
# 2. the imagestream was observed before the image creation, i.e.
# this record was created recently and it should be protected by keep_younger_than
continue
image = self.image_mapping[item["image"]]
# check prune over limit size
if prune_over_size_limit and not self.exceeds_limits(
stream_namespace, image
):
filtered_items.append(item)
continue
image_ref = "%s/%s@%s" % (stream_namespace, stream_name, item["image"])
if image_ref in self.used_images:
# keeping because tag is used
filtered_items.append(item)
continue
images_to_delete.append(item["image"])
if self.params.get("prune_registry"):
manifests_to_delete.append(image["metadata"]["name"])
path = stream_namespace + "/" + stream_name
image_blobs, err = get_image_blobs(image)
if not err:
self.delete_layers_links(path, image_blobs)
return filtered_items, manifests_to_delete, images_to_delete
def prune_image_streams(self, stream):
name = stream["metadata"]["namespace"] + "/" + stream["metadata"]["name"]
if is_too_young_object(stream, self.max_creation_timestamp):
# keeping all images because of image stream too young
return None, []
facts = self.kubernetes_facts(
kind="ImageStream",
api_version=ApiConfiguration.get("ImageStream"),
name=stream["metadata"]["name"],
namespace=stream["metadata"]["namespace"],
)
image_stream = facts.get("resources")
if len(image_stream) != 1:
# skipping because it does not exist anymore
return None, []
stream = image_stream[0]
namespace = self.params.get("namespace")
stream_to_update = not namespace or (
stream["metadata"]["namespace"] == namespace
)
manifests_to_delete, images_to_delete = [], []
deleted_items = False
# Update Image stream tag
if stream_to_update:
tags = stream["status"].get("tags", [])
for idx, tag_event_list in enumerate(tags):
(
filtered_tag_event,
tag_manifests_to_delete,
tag_images_to_delete,
) = self.prune_image_stream_tag(stream, tag_event_list)
stream["status"]["tags"][idx]["items"] = filtered_tag_event
manifests_to_delete += tag_manifests_to_delete
images_to_delete += tag_images_to_delete
deleted_items = deleted_items or (len(tag_images_to_delete) > 0)
# Deleting tags without items
tags = []
for tag in stream["status"].get("tags", []):
if tag["items"] is None or len(tag["items"]) == 0:
continue
tags.append(tag)
stream["status"]["tags"] = tags
result = None
# Update ImageStream
if stream_to_update:
if deleted_items:
result = self.update_image_stream_status(stream)
if self.params.get("prune_registry"):
self.delete_manifests(name, manifests_to_delete)
return result, images_to_delete
def prune_images(self, image):
if not self.params.get("all_images"):
if read_object_annotation(image, "openshift.io/image.managed") != "true":
# keeping external image because all_images is set to false
# pruning only managed images
return None
if is_too_young_object(image, self.max_creation_timestamp):
# keeping because of keep_younger_than
return None
# Deleting image from registry
if self.params.get("prune_registry"):
image_blobs, err = get_image_blobs(image)
if err:
self.fail_json(msg=err)
# add blob for image name
image_blobs.append(image["metadata"]["name"])
self.delete_blobs(image_blobs)
# Delete image from cluster
return self.delete_image(image)
def execute_module(self):
resources = self.list_objects()
if not self.check_mode and self.params.get("prune_registry"):
if not self.registryhost:
self.registryhost = determine_host_registry(
self.module, resources["Image"], resources["ImageStream"]
)
# validate that host has a scheme
if "://" not in self.registryhost:
self.registryhost = "https://" + self.registryhost
# Analyze Image Streams
analyze_ref = OpenShiftAnalyzeImageStream(
ignore_invalid_refs=self.params.get("ignore_invalid_refs"),
max_creation_timestamp=self.max_creation_timestamp,
module=self.module,
)
self.used_tags, self.used_images, error = analyze_ref.analyze_image_stream(
resources
)
if error:
self.fail_json(msg=error)
# Create image mapping
self.image_mapping = {}
for m in resources["Image"]:
self.image_mapping[m["metadata"]["name"]] = m
# Create limit range mapping
self.limit_range = {}
for limit in resources["LimitRange"]:
namespace = limit["metadata"]["namespace"]
if namespace not in self.limit_range:
self.limit_range[namespace] = []
self.limit_range[namespace].append(limit)
# Stage 1: delete history from image streams
updated_image_streams = []
deleted_tags_images = []
updated_is_mapping = {}
for stream in resources["ImageStream"]:
result, images_to_delete = self.prune_image_streams(stream)
if result:
updated_is_mapping[
result["metadata"]["namespace"] + "/" + result["metadata"]["name"]
] = result
updated_image_streams.append(result)
deleted_tags_images += images_to_delete
# Create a list with images referenced on image stream
self.referenced_images = []
for item in self.kubernetes_facts(
kind="ImageStream", api_version="image.openshift.io/v1"
)["resources"]:
name = "%s/%s" % (item["metadata"]["namespace"], item["metadata"]["name"])
if name in updated_is_mapping:
item = updated_is_mapping[name]
for tag in item["status"].get("tags", []):
self.referenced_images += [t["image"] for t in tag["items"] or []]
# Stage 2: delete images
images = []
images_to_delete = [x["metadata"]["name"] for x in resources["Image"]]
if self.params.get("namespace") is not None:
# When namespace is defined, prune only images that were referenced by ImageStream
# from the corresponding namespace
images_to_delete = deleted_tags_images
for name in images_to_delete:
if name in self.referenced_images:
# The image is referenced in one or more Image stream
continue
if name not in self.image_mapping:
# The image is not existing anymore
continue
result = self.prune_images(self.image_mapping[name])
if result:
images.append(result)
result = {
"changed": self.changed,
"deleted_images": images,
"updated_image_streams": updated_image_streams,
}
self.exit_json(**result)