summaryrefslogtreecommitdiffstats
path: root/roles/openshift_health_checker/openshift_checks/docker_image_availability.py
diff options
context:
space:
mode:
Diffstat (limited to 'roles/openshift_health_checker/openshift_checks/docker_image_availability.py')
-rw-r--r--roles/openshift_health_checker/openshift_checks/docker_image_availability.py240
1 files changed, 240 insertions, 0 deletions
diff --git a/roles/openshift_health_checker/openshift_checks/docker_image_availability.py b/roles/openshift_health_checker/openshift_checks/docker_image_availability.py
new file mode 100644
index 000000000..7c8ac78fe
--- /dev/null
+++ b/roles/openshift_health_checker/openshift_checks/docker_image_availability.py
@@ -0,0 +1,240 @@
+"""Check that required Docker images are available."""
+
+from pipes import quote
+from ansible.module_utils import six
+from openshift_checks import OpenShiftCheck
+from openshift_checks.mixins import DockerHostMixin
+
+
+NODE_IMAGE_SUFFIXES = ["haproxy-router", "docker-registry", "deployer", "pod"]
+DEPLOYMENT_IMAGE_INFO = {
+ "origin": {
+ "namespace": "openshift",
+ "name": "origin",
+ "registry_console_image": "cockpit/kubernetes",
+ },
+ "openshift-enterprise": {
+ "namespace": "openshift3",
+ "name": "ose",
+ "registry_console_image": "registry.access.redhat.com/openshift3/registry-console",
+ },
+}
+
+
+class DockerImageAvailability(DockerHostMixin, OpenShiftCheck):
+ """Check that required Docker images are available.
+
+ Determine docker images that an install would require and check that they
+ are either present in the host's docker index, or available for the host to pull
+ with known registries as defined in our inventory file (or defaults).
+ """
+
+ name = "docker_image_availability"
+ tags = ["preflight"]
+ # we use python-docker-py to check local docker for images, and skopeo
+ # to look for images available remotely without waiting to pull them.
+ dependencies = ["python-docker-py", "skopeo"]
+ # command for checking if remote registries have an image, without docker pull
+ skopeo_command = "timeout 10 skopeo inspect --tls-verify={tls} {creds} docker://{registry}/{image}"
+ skopeo_example_command = "skopeo inspect [--tls-verify=false] [--creds=<user>:<pass>] docker://<registry>/<image>"
+
+ def __init__(self, *args, **kwargs):
+ super(DockerImageAvailability, self).__init__(*args, **kwargs)
+
+ self.registries = dict(
+ # set of registries that need to be checked insecurely (note: not accounting for CIDR entries)
+ insecure=set(self.ensure_list("openshift_docker_insecure_registries")),
+ # set of registries that should never be queried even if given in the image
+ blocked=set(self.ensure_list("openshift_docker_blocked_registries")),
+ )
+
+ # ordered list of registries (according to inventory vars) that docker will try for unscoped images
+ regs = self.ensure_list("openshift_docker_additional_registries")
+ # currently one of these registries is added whether the user wants it or not.
+ deployment_type = self.get_var("openshift_deployment_type")
+ if deployment_type == "origin" and "docker.io" not in regs:
+ regs.append("docker.io")
+ elif deployment_type == 'openshift-enterprise' and "registry.access.redhat.com" not in regs:
+ regs.append("registry.access.redhat.com")
+ self.registries["configured"] = regs
+
+ # for the oreg_url registry there may be credentials specified
+ components = self.get_var("oreg_url", default="").split('/')
+ self.registries["oreg"] = "" if len(components) < 3 else components[0]
+ self.skopeo_command_creds = ""
+ oreg_auth_user = self.get_var('oreg_auth_user', default='')
+ oreg_auth_password = self.get_var('oreg_auth_password', default='')
+ if oreg_auth_user != '' and oreg_auth_password != '':
+ self.skopeo_command_creds = "--creds={}:{}".format(quote(oreg_auth_user), quote(oreg_auth_password))
+
+ # record whether we could reach a registry or not (and remember results)
+ self.reachable_registries = {}
+
+ def is_active(self):
+ """Skip hosts with unsupported deployment types."""
+ deployment_type = self.get_var("openshift_deployment_type")
+ has_valid_deployment_type = deployment_type in DEPLOYMENT_IMAGE_INFO
+
+ return super(DockerImageAvailability, self).is_active() and has_valid_deployment_type
+
+ def run(self):
+ msg, failed = self.ensure_dependencies()
+ if failed:
+ return {
+ "failed": True,
+ "msg": "Some dependencies are required in order to check Docker image availability.\n" + msg
+ }
+
+ required_images = self.required_images()
+ missing_images = set(required_images) - set(self.local_images(required_images))
+
+ # exit early if all images were found locally
+ if not missing_images:
+ return {}
+
+ available_images = self.available_images(missing_images)
+ unavailable_images = set(missing_images) - set(available_images)
+
+ if unavailable_images:
+ unreachable = [reg for reg, reachable in self.reachable_registries.items() if not reachable]
+ unreachable_msg = "Failed connecting to: {}\n".format(", ".join(unreachable))
+ blocked_msg = "Blocked registries: {}\n".format(", ".join(self.registries["blocked"]))
+ msg = (
+ "One or more required container images are not available:\n {missing}\n"
+ "Checked with: {cmd}\n"
+ "Default registries searched: {registries}\n"
+ "{blocked}"
+ "{unreachable}"
+ ).format(
+ missing=",\n ".join(sorted(unavailable_images)),
+ cmd=self.skopeo_example_command,
+ registries=", ".join(self.registries["configured"]),
+ blocked=blocked_msg if self.registries["blocked"] else "",
+ unreachable=unreachable_msg if unreachable else "",
+ )
+
+ return dict(failed=True, msg=msg)
+
+ return {}
+
+ def required_images(self):
+ """
+ Determine which images we expect to need for this host.
+ Returns: a set of required images like 'openshift/origin:v3.6'
+
+ The thorny issue of determining the image names from the variables is under consideration
+ via https://github.com/openshift/openshift-ansible/issues/4415
+
+ For now we operate as follows:
+ * For containerized components (master, node, ...) we look at the deployment type and
+ use openshift/origin or openshift3/ose as the base for those component images. The
+ version is openshift_image_tag as determined by the openshift_version role.
+ * For OpenShift-managed infrastructure (router, registry...) we use oreg_url if
+ it is defined; otherwise we again use the base that depends on the deployment type.
+ Registry is not included in constructed images. It may be in oreg_url or etcd image.
+ """
+ required = set()
+ deployment_type = self.get_var("openshift_deployment_type")
+ host_groups = self.get_var("group_names")
+ # containerized etcd may not have openshift_image_tag, see bz 1466622
+ image_tag = self.get_var("openshift_image_tag", default="latest")
+ image_info = DEPLOYMENT_IMAGE_INFO[deployment_type]
+
+ # template for images that run on top of OpenShift
+ image_url = "{}/{}-{}:{}".format(image_info["namespace"], image_info["name"], "${component}", "${version}")
+ image_url = self.get_var("oreg_url", default="") or image_url
+ if 'oo_nodes_to_config' in host_groups:
+ for suffix in NODE_IMAGE_SUFFIXES:
+ required.add(image_url.replace("${component}", suffix).replace("${version}", image_tag))
+ # The registry-console is for some reason not prefixed with ose- like the other components.
+ # Nor is it versioned the same, so just look for latest.
+ # Also a completely different name is used for Origin.
+ required.add(image_info["registry_console_image"])
+
+ # images for containerized components
+ if self.get_var("openshift", "common", "is_containerized"):
+ components = set()
+ if 'oo_nodes_to_config' in host_groups:
+ components.update(["node", "openvswitch"])
+ if 'oo_masters_to_config' in host_groups: # name is "origin" or "ose"
+ components.add(image_info["name"])
+ for component in components:
+ required.add("{}/{}:{}".format(image_info["namespace"], component, image_tag))
+ if 'oo_etcd_to_config' in host_groups: # special case, note it is the same for origin/enterprise
+ required.add("registry.access.redhat.com/rhel7/etcd") # and no image tag
+
+ return required
+
+ def local_images(self, images):
+ """Filter a list of images and return those available locally."""
+ found_images = []
+ for image in images:
+ # docker could have the image name as-is or prefixed with any registry
+ imglist = [image] + [reg + "/" + image for reg in self.registries["configured"]]
+ if self.is_image_local(imglist):
+ found_images.append(image)
+ return found_images
+
+ def is_image_local(self, image):
+ """Check if image is already in local docker index."""
+ result = self.execute_module("docker_image_facts", {"name": image})
+ return bool(result.get("images")) and not result.get("failed")
+
+ def ensure_list(self, registry_param):
+ """Return the task var as a list."""
+ # https://bugzilla.redhat.com/show_bug.cgi?id=1497274
+ # If the result was a string type, place it into a list. We must do this
+ # as using list() on a string will split the string into its characters.
+ # Otherwise cast to a list as was done previously.
+ registry = self.get_var(registry_param, default=[])
+ if not isinstance(registry, six.string_types):
+ return list(registry)
+ return self.normalize(registry)
+
+ def available_images(self, images):
+ """Search remotely for images. Returns: list of images found."""
+ return [
+ image for image in images
+ if self.is_available_skopeo_image(image)
+ ]
+
+ def is_available_skopeo_image(self, image):
+ """Use Skopeo to determine if required image exists in known registry(s)."""
+ registries = self.registries["configured"]
+ # If image already includes a registry, only use that.
+ # NOTE: This logic would incorrectly identify images that do not use a namespace, e.g.
+ # registry.access.redhat.com/rhel7 as if the registry were a namespace.
+ # It's not clear that there's any way to distinguish them, but fortunately
+ # the current set of images all look like [registry/]namespace/name[:version].
+ if image.count("/") > 1:
+ registry, image = image.split("/", 1)
+ registries = [registry]
+
+ for registry in registries:
+ if registry in self.registries["blocked"]:
+ continue # blocked will never be consulted
+ if registry not in self.reachable_registries:
+ self.reachable_registries[registry] = self.connect_to_registry(registry)
+ if not self.reachable_registries[registry]:
+ continue # do not keep trying unreachable registries
+
+ args = dict(registry=registry, image=image)
+ args["tls"] = "false" if registry in self.registries["insecure"] else "true"
+ args["creds"] = self.skopeo_command_creds if registry == self.registries["oreg"] else ""
+
+ result = self.execute_module_with_retries("command", {"_raw_params": self.skopeo_command.format(**args)})
+ if result.get("rc", 0) == 0 and not result.get("failed"):
+ return True
+ if result.get("rc") == 124: # RC 124 == timed out; mark unreachable
+ self.reachable_registries[registry] = False
+
+ return False
+
+ def connect_to_registry(self, registry):
+ """Use ansible wait_for module to test connectivity from host to registry. Returns bool."""
+ # test a simple TCP connection
+ host, _, port = registry.partition(":")
+ port = port or 443
+ args = dict(host=host, port=port, state="started", timeout=30)
+ result = self.execute_module("wait_for", args)
+ return result.get("rc", 0) == 0 and not result.get("failed")