def create_config(self):
"""
Create the OCP deploy config, if something needs to be changed fro
specific platform you can overload this method in child class.
"""
# Generate install-config from template
logger.info("Generating install-config")
_templating = templating.Templating()
ocp_install_template = (f"install-config-{self.deployment_platform}-"
f"{self.deployment_type}.yaml.j2")
ocp_install_template_path = os.path.join("ocp-deployment",
ocp_install_template)
install_config_str = _templating.render_template(
ocp_install_template_path, config.ENV_DATA)
# Log the install config *before* adding the pull secret,
# so we don't leak sensitive data.
logger.info(f"Install config: \n{install_config_str}")
# Parse the rendered YAML so that we can manipulate the object directly
install_config_obj = yaml.safe_load(install_config_str)
install_config_obj['pullSecret'] = self.get_pull_secret()
ssh_key = self.get_ssh_key()
if ssh_key:
install_config_obj['sshKey'] = ssh_key
install_config_str = yaml.safe_dump(install_config_obj)
install_config = os.path.join(self.cluster_path, "install-config.yaml")
with open(install_config, "w") as f:
f.write(install_config_str)
def build_ansible_inventory(self, hosts):
"""
Build the ansible hosts file from jinja template
Args:
hosts (list): list of private host names
Returns:
str: path of the ansible file created
"""
_templating = templating.Templating()
ansible_host_file = dict()
ansible_host_file["ansible_user"] = "******"
ansible_host_file["ansible_become"] = "True"
ansible_host_file["pod_kubeconfig"] = "/kubeconfig"
ansible_host_file["pod_pull_secret"] = "/tmp/pull-secret"
ansible_host_file["rhel_worker_nodes"] = hosts
logging.info(ansible_host_file)
data = _templating.render_template(
constants.ANSIBLE_INVENTORY_YAML,
ansible_host_file,
)
logging.debug("Ansible hosts file:%s", data)
host_file_path = "/tmp/hosts"
with open(host_file_path, "w") as f:
f.write(data)
return host_file_path
def deploy_ocp_prereq(self):
"""
Perform generic prereq before calling openshift-installer
This method performs all the basic steps necessary before invoking the
installer
"""
if self.teardown and not self.deploy:
msg = f"Attempting teardown of non-accessible cluster: "
msg += f"{self.cluster_path}"
pytest.fail(msg)
elif not self.deploy and not self.teardown:
msg = "The given cluster can not be connected to: {}. ".format(
self.cluster_path)
msg += (f"Provide a valid --cluster-path or use --deploy to "
f"deploy a new cluster")
pytest.fail(msg)
elif not system.is_path_empty(self.cluster_path) and self.deploy:
msg = "The given cluster path is not empty: {}. ".format(
self.cluster_path)
msg += (f"Provide an empty --cluster-path and --deploy to deploy "
f"a new cluster")
pytest.fail(msg)
else:
logger.info(
f"A testing cluster will be deployed and cluster information "
f"stored at: %s", self.cluster_path)
# Generate install-config from template
logger.info("Generating install-config")
pull_secret_path = os.path.join(constants.TOP_DIR, "data",
"pull-secret")
_templating = templating.Templating()
install_config_str = _templating.render_template(
"install-config.yaml.j2", config.ENV_DATA)
# Log the install config *before* adding the pull secret,
# so we don't leak sensitive data.
logger.info(f"Install config: \n{install_config_str}")
# Parse the rendered YAML so that we can manipulate the object directly
install_config_obj = yaml.safe_load(install_config_str)
with open(pull_secret_path, "r") as f:
# Parse, then unparse, the JSON file.
# We do this for two reasons: to ensure it is well-formatted, and
# also to ensure it ends up as a single line.
install_config_obj['pullSecret'] = json.dumps(json.loads(f.read()))
install_config_str = yaml.safe_dump(install_config_obj)
install_config = os.path.join(self.cluster_path, "install-config.yaml")
with open(install_config, "w") as f:
f.write(install_config_str)
def cleanup(cluster_name, cluster_id):
"""
Cleanup existing cluster in AWS
Args:
cluster_name (str): Name of the cluster
cluster_id (str): Cluster id to cleanup
"""
data = {'cluster_name': cluster_name, 'cluster_id': cluster_id}
template = templating.Templating(base_path=TEMPLATE_CLEANUP_DIR)
cleanup_template = template.render_template(CLEANUP_YAML, data)
cleanup_path = tempfile.mkdtemp(prefix='cleanup_')
cleanup_file = os.path.join(cleanup_path, 'metadata.json')
with open(cleanup_file, "w") as temp:
temp.write(cleanup_template)
bin_dir = os.path.expanduser(config.RUN['bin_dir'])
oc_bin = os.path.join(bin_dir, "openshift-install")
logger.info(f"cleaning up {cluster_id}")
run_cmd(f"{oc_bin} destroy cluster --dir {cleanup_path} --log-level=debug")
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
_templating = templating.Templating()
ceph_cluster = ocp.OCP(
kind='CephCluster', namespace=self.namespace
)
try:
ceph_cluster.get().get('items')[0]
logger.warning("OCS cluster already exists")
return
except (IndexError, CommandFailed):
logger.info("Running OCS basic installation")
if not self.ocs_operator_deployment:
create_oc_resource(
'common.yaml', self.cluster_path, _templating, config.ENV_DATA
)
run_cmd(
f'oc label namespace {config.ENV_DATA["cluster_namespace"]} '
f'"openshift.io/cluster-monitoring=true"'
)
run_cmd(
f"oc policy add-role-to-user view "
f"system:serviceaccount:openshift-monitoring:prometheus-k8s "
f"-n {self.namespace}"
)
# HACK: If you would like to drop this hack, make sure that you
# also updated docs and write appropriate unit/integration tests
# for config processing.
if config.ENV_DATA.get('monitoring_enabled') in (
"true", "True", True
):
# RBAC rules for monitoring, based on documentation change in
# rook:
# https://github.com/rook/rook/commit/1b6fe840f6ae7372a9675ba727ecc65326708aa8
# HACK: This should be dropped when OCS is managed by OLM
apply_oc_resource(
'rbac.yaml',
self.cluster_path,
_templating,
config.ENV_DATA,
template_dir="monitoring"
)
# Increased to 15 seconds as 10 is not enough
# TODO: do the sampler function and check if resource exist
wait_time = 15
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
create_oc_resource(
'operator-openshift.yaml', self.cluster_path,
_templating, config.ENV_DATA
)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
run_cmd(
f"oc wait --for condition=ready pod "
f"-l app=rook-ceph-operator "
f"-n {self.namespace} "
f"--timeout=120s"
)
run_cmd(
f"oc wait --for condition=ready pod "
f"-l app=rook-discover "
f"-n {self.namespace} "
f"--timeout=120s"
)
create_oc_resource(
'cluster.yaml', self.cluster_path, _templating, config.ENV_DATA
)
else:
self.deploy_ocs_via_operator()
pod = ocp.OCP(
kind=constants.POD, namespace=self.namespace
)
cfs = ocp.OCP(
kind=constants.CEPHFILESYSTEM,
namespace=self.namespace
)
# Check for Ceph pods
assert pod.wait_for_resource(
condition='Running', selector='app=rook-ceph-mon',
resource_count=3, timeout=600
)
assert pod.wait_for_resource(
condition='Running', selector='app=rook-ceph-mgr',
timeout=600
)
assert pod.wait_for_resource(
condition='Running', selector='app=rook-ceph-osd',
resource_count=3, timeout=600
)
# validate ceph mon/osd volumes are backed by pvc
validate_cluster_on_pvc()
# Creating toolbox pod
setup_ceph_toolbox()
assert pod.wait_for_resource(
condition=constants.STATUS_RUNNING,
selector='app=rook-ceph-tools', resource_count=1, timeout=600
)
if not self.ocs_operator_deployment:
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
# HACK: This should be dropped (including service-monitor.yaml and
# prometheus-rules.yaml files) when OCS is managed by OLM
if config.ENV_DATA.get('monitoring_enabled') not in (
"true", "True", True
):
# HACK: skip creation of rook-ceph-mgr service monitor when
# monitoring is enabled (if this were not skipped, the step
# would fail because rook would create the service monitor at
# this point already)
create_oc_resource(
"service-monitor.yaml", self.cluster_path, _templating,
config.ENV_DATA
)
# HACK: skip creation of prometheus-rules, rook-ceph is
# concerned with it's setup now, based on clarification from
# Umanga Chapagain
create_oc_resource(
"prometheus-rules.yaml", self.cluster_path, _templating,
config.ENV_DATA
)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
# Create MDS pods for CephFileSystem
fs_data = templating.load_yaml(constants.CEPHFILESYSTEM_YAML)
fs_data['metadata']['namespace'] = self.namespace
ceph_obj = OCS(**fs_data)
ceph_obj.create()
assert pod.wait_for_resource(
condition=constants.STATUS_RUNNING, selector='app=rook-ceph-mds',
resource_count=2, timeout=600
)
# Check for CephFilesystem creation in ocp
cfs_data = cfs.get()
cfs_name = cfs_data['items'][0]['metadata']['name']
if helpers.validate_cephfilesystem(cfs_name):
logger.info(f"MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
logger.error(
f"MDS deployment Failed! Please check logs!"
)
if config.ENV_DATA.get('monitoring_enabled') and config.ENV_DATA.get('persistent-monitoring'):
# Create a pool, secrets and sc
secret_obj = helpers.create_secret(interface_type=constants.CEPHBLOCKPOOL)
cbj_obj = helpers.create_ceph_block_pool()
sc_obj = helpers.create_storage_class(
interface_type=constants.CEPHBLOCKPOOL,
interface_name=cbj_obj.name,
secret_name=secret_obj.name
)
# Get the list of monitoring pods
pods_list = get_all_pods(
namespace=defaults.OCS_MONITORING_NAMESPACE,
selector=['prometheus', 'alertmanager']
)
# Create configmap cluster-monitoring-config
create_configmap_cluster_monitoring_pod(sc_obj.name)
# Take some time to respin the pod
waiting_time = 30
logger.info(f"Waiting {waiting_time} seconds...")
time.sleep(waiting_time)
# Validate the pods are respinned and in running state
validate_pods_are_respinned_and_running_state(
pods_list
)
# Validate the pvc is created on monitoring pods
validate_pvc_created_and_bound_on_monitoring_pods()
# Validate the pvc are mounted on pods
validate_pvc_are_mounted_on_monitoring_pods(pods_list)
# Change registry backend to OCS CEPHFS RWX PVC
registry.change_registry_backend_to_ocs()
# Verify health of ceph cluster
# TODO: move destroy cluster logic to new CLI usage pattern?
logger.info("Done creating rook resources, waiting for HEALTH_OK")
assert ceph_health_check(
namespace=self.namespace
)
# patch gp2/thin storage class as 'non-default'
self.patch_default_sc_to_non_default()
size is returned to backend pool
"""
import logging
import pytest
from ocs_ci.ocs import constants
from ocs_ci.ocs.exceptions import UnexpectedBehaviour
from tests import helpers
from ocs_ci.framework.testlib import tier1, acceptance, ManageTest
from ocs_ci.utility import templating
from ocs_ci.utility.retry import retry
from ocs_ci.ocs.resources import pod
logger = logging.getLogger(__name__)
_templating = templating.Templating()
@retry(UnexpectedBehaviour, tries=5, delay=3, backoff=1)
def verify_pv_not_exists(pvc_obj, cbp_name, rbd_image_id):
"""
Ensure that pv does not exists
"""
# Validate on ceph side
logger.info(f"Verifying PV {pvc_obj.backed_pv} exists on backend")
status = helpers.verify_volume_deleted_in_backend(
interface=constants.CEPHBLOCKPOOL,
image_uuid=rbd_image_id,
pool_name=cbp_name)
def cleanup(cluster_name, cluster_id, upi=False, failed_deletions=None):
"""
Cleanup existing cluster in AWS
Args:
cluster_name (str): Name of the cluster
cluster_id (str): Cluster id to cleanup
upi (bool): True for UPI cluster, False otherwise
failed_deletions (list): list of clusters we failed to delete, used
for reporting purposes
"""
data = {'cluster_name': cluster_name, 'cluster_id': cluster_id}
template = templating.Templating(base_path=TEMPLATE_CLEANUP_DIR)
cleanup_template = template.render_template(CLEANUP_YAML, data)
cleanup_path = tempfile.mkdtemp(prefix='cleanup_')
cleanup_file = os.path.join(cleanup_path, 'metadata.json')
with open(cleanup_file, "w") as temp:
temp.write(cleanup_template)
bin_dir = os.path.expanduser(config.RUN['bin_dir'])
oc_bin = os.path.join(bin_dir, "openshift-install")
if upi:
aws = AWS()
rhel_workers = get_rhel_worker_instances(cleanup_path)
logger.info(f"{cluster_name}'s RHEL workers: {rhel_workers}")
if rhel_workers:
terminate_rhel_workers(rhel_workers)
# Destroy extra volumes
destroy_volumes(cluster_name)
stack_names = list()
# Get master, bootstrap and security group stacks
for stack_type in ['ma', 'bs', 'sg']:
try:
stack_names.append(
aws.get_cloudformation_stacks(
pattern=f"{cluster_name}-{stack_type}")[0]
['StackName'])
except ClientError:
continue
# Get the worker stacks
worker_index = 0
worker_stack_exists = True
while worker_stack_exists:
try:
stack_names.append(
aws.get_cloudformation_stacks(
pattern=f"{cluster_name}-no{worker_index}")[0]
['StackName'])
worker_index += 1
except ClientError:
worker_stack_exists = False
logger.info(f"Deleting stacks: {stack_names}")
aws.delete_cloudformation_stacks(stack_names)
# Destroy the cluster
logger.info(f"cleaning up {cluster_id}")
destroy_cluster(installer=oc_bin, cluster_path=cleanup_path)
for stack_type in ['inf', 'vpc']:
try:
stack_names.append(
aws.get_cloudformation_stacks(
pattern=f"{cluster_name}-{stack_type}")[0]
['StackName'])
except ClientError:
continue
try:
aws.delete_cloudformation_stacks(stack_names)
except StackStatusError:
logger.error('Failed to fully destroy cluster %s', cluster_name)
if failed_deletions:
failed_deletions.append(cluster_name)
raise
else:
logger.info(f"cleaning up {cluster_id}")
try:
destroy_cluster(installer=oc_bin, cluster_path=cleanup_path)
except CommandFailed:
logger.error('Failed to fully destroy cluster %s', cluster_name)
if failed_deletions:
failed_deletions.append(cluster_name)
raise
def cluster(request):
log.info(f"All logs located at {log_path}")
log.info("Running OCS basic installation")
cluster_path = config.ENV_DATA['cluster_path']
deploy = config.RUN['cli_params']['deploy']
teardown = config.RUN['cli_params']['teardown']
# Add a finalizer to teardown the cluster after test execution is finished
if teardown:
request.addfinalizer(cluster_teardown)
log.info("Will teardown cluster because --teardown was provided")
# Test cluster access and if exist just skip the deployment.
if is_cluster_running(cluster_path):
log.info("The installation is skipped because the cluster is running")
return
elif teardown and not deploy:
log.info("Attempting teardown of non-accessible cluster: %s",
cluster_path)
return
elif not deploy and not teardown:
msg = "The given cluster can not be connected to: {}. ".format(
cluster_path)
msg += "Provide a valid --cluster-path or use --deploy to deploy a new cluster"
pytest.fail(msg)
elif not system.is_path_empty(cluster_path) and deploy:
msg = "The given cluster path is not empty: {}. ".format(cluster_path)
msg += "Provide an empty --cluster-path and --deploy to deploy a new cluster"
pytest.fail(msg)
else:
log.info(
"A testing cluster will be deployed and cluster information stored at: %s",
cluster_path)
# Generate install-config from template
log.info("Generating install-config")
pull_secret_path = os.path.join(constants.TOP_DIR, "data", "pull-secret")
# TODO: check for supported platform and raise the exception if not
# supported. Currently we support just AWS.
_templating = templating.Templating()
install_config_str = _templating.render_template("install-config.yaml.j2",
config.ENV_DATA)
# Log the install config *before* adding the pull secret, so we don't leak
# sensitive data.
log.info(f"Install config: \n{install_config_str}")
# Parse the rendered YAML so that we can manipulate the object directly
install_config_obj = yaml.safe_load(install_config_str)
with open(pull_secret_path, "r") as f:
# Parse, then unparse, the JSON file.
# We do this for two reasons: to ensure it is well-formatted, and
# also to ensure it ends up as a single line.
install_config_obj['pullSecret'] = json.dumps(json.loads(f.read()))
install_config_str = yaml.safe_dump(install_config_obj)
install_config = os.path.join(cluster_path, "install-config.yaml")
with open(install_config, "w") as f:
f.write(install_config_str)
# Download installer
installer = get_openshift_installer(config.DEPLOYMENT['installer_version'])
# Download client
get_openshift_client()
# Deploy cluster
log.info("Deploying cluster")
run_cmd(f"{installer} create cluster "
f"--dir {cluster_path} "
f"--log-level debug")
# Test cluster access
if not OCP.set_kubeconfig(
os.path.join(cluster_path, config.RUN.get('kubeconfig_location'))):
pytest.fail("Cluster is not available!")
# TODO: Create cluster object, add to config.ENV_DATA for other tests to
# utilize.
# Determine worker pattern and create ebs volumes
with open(os.path.join(cluster_path, "terraform.tfvars")) as f:
tfvars = json.load(f)
cluster_id = tfvars['cluster_id']
worker_pattern = f'{cluster_id}-worker*'
log.info(f'Worker pattern: {worker_pattern}')
create_ebs_volumes(worker_pattern, region_name=config.ENV_DATA['region'])
# render templates and create resources
create_oc_resource('common.yaml', cluster_path, _templating,
config.ENV_DATA)
run_cmd(f'oc label namespace {config.ENV_DATA["cluster_namespace"]} '
f'"openshift.io/cluster-monitoring=true"')
run_cmd(f"oc policy add-role-to-user view "
f"system:serviceaccount:openshift-monitoring:prometheus-k8s "
f"-n {config.ENV_DATA['cluster_namespace']}")
apply_oc_resource('csi-nodeplugin-rbac_rbd.yaml',
cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/rbd/")
apply_oc_resource('csi-provisioner-rbac_rbd.yaml',
cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/rbd/")
apply_oc_resource('csi-nodeplugin-rbac_cephfs.yaml',
cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/cephfs/")
apply_oc_resource('csi-provisioner-rbac_cephfs.yaml',
cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/cephfs/")
# Increased to 15 seconds as 10 is not enough
# TODO: do the sampler function and check if resource exist
wait_time = 15
log.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
create_oc_resource('operator-openshift-with-csi.yaml', cluster_path,
_templating, config.ENV_DATA)
log.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
run_cmd(f"oc wait --for condition=ready pod "
f"-l app=rook-ceph-operator "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s")
run_cmd(f"oc wait --for condition=ready pod "
f"-l app=rook-discover "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s")
create_oc_resource('cluster.yaml', cluster_path, _templating,
config.ENV_DATA)
POD = ocp.OCP(kind=constants.POD,
namespace=config.ENV_DATA['cluster_namespace'])
CFS = ocp.OCP(kind=constants.CEPHFILESYSTEM,
namespace=config.ENV_DATA['cluster_namespace'])
# Check for the Running status of Ceph Pods
run_cmd(f"oc wait --for condition=ready pod "
f"-l app=rook-ceph-agent "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s")
assert POD.wait_for_resource(condition='Running',
selector='app=rook-ceph-mon',
resource_count=3,
timeout=600)
assert POD.wait_for_resource(condition='Running',
selector='app=rook-ceph-mgr',
timeout=600)
assert POD.wait_for_resource(condition='Running',
selector='app=rook-ceph-osd',
resource_count=3,
timeout=600)
create_oc_resource('toolbox.yaml', cluster_path, _templating,
config.ENV_DATA)
log.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
create_oc_resource('storage-manifest.yaml', cluster_path, _templating,
config.ENV_DATA)
create_oc_resource("service-monitor.yaml", cluster_path, _templating,
config.ENV_DATA)
create_oc_resource("prometheus-rules.yaml", cluster_path, _templating,
config.ENV_DATA)
log.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
# Create MDS pods for CephFileSystem
fs_data = templating.load_yaml_to_dict(constants.CEPHFILESYSTEM_YAML)
fs_data['metadata']['namespace'] = config.ENV_DATA['cluster_namespace']
ceph_obj = OCS(**fs_data)
ceph_obj.create()
assert POD.wait_for_resource(condition=constants.STATUS_RUNNING,
selector='app=rook-ceph-mds',
resource_count=2,
timeout=600)
# Check for CephFilesystem creation in ocp
cfs_data = CFS.get()
cfs_name = cfs_data['items'][0]['metadata']['name']
if helpers.validate_cephfilesystem(cfs_name):
log.info(f"MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
log.error(f"MDS deployment Failed! Please check logs!")
# Verify health of ceph cluster
# TODO: move destroy cluster logic to new CLI usage pattern?
log.info("Done creating rook resources, waiting for HEALTH_OK")
assert ceph_health_check(namespace=config.ENV_DATA['cluster_namespace'])
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
_templating = templating.Templating()
ceph_cluster = ocp.OCP(kind='CephCluster',
namespace=config.ENV_DATA['cluster_namespace'])
try:
ceph_cluster.get().get('items')[0]
logger.warning("OCS cluster already exists")
return
except (IndexError, CommandFailed):
logger.info("Running OCS basic installation")
create_oc_resource('common.yaml', self.cluster_path, _templating,
config.ENV_DATA)
run_cmd(f'oc label namespace {config.ENV_DATA["cluster_namespace"]} '
f'"openshift.io/cluster-monitoring=true"')
run_cmd(f"oc policy add-role-to-user view "
f"system:serviceaccount:openshift-monitoring:prometheus-k8s "
f"-n {config.ENV_DATA['cluster_namespace']}")
apply_oc_resource('csi-nodeplugin-rbac_rbd.yaml',
self.cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/rbd/")
apply_oc_resource('csi-provisioner-rbac_rbd.yaml',
self.cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/rbd/")
apply_oc_resource('csi-nodeplugin-rbac_cephfs.yaml',
self.cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/cephfs/")
apply_oc_resource('csi-provisioner-rbac_cephfs.yaml',
self.cluster_path,
_templating,
config.ENV_DATA,
template_dir="ocs-deployment/csi/cephfs/")
# Increased to 15 seconds as 10 is not enough
# TODO: do the sampler function and check if resource exist
wait_time = 15
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
create_oc_resource('operator-openshift-with-csi.yaml',
self.cluster_path, _templating, config.ENV_DATA)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
run_cmd(f"oc wait --for condition=ready pod "
f"-l app=rook-ceph-operator "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s")
run_cmd(f"oc wait --for condition=ready pod "
f"-l app=rook-discover "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s")
create_oc_resource('cluster.yaml', self.cluster_path, _templating,
config.ENV_DATA)
pod = ocp.OCP(kind=constants.POD,
namespace=config.ENV_DATA['cluster_namespace'])
cfs = ocp.OCP(kind=constants.CEPHFILESYSTEM,
namespace=config.ENV_DATA['cluster_namespace'])
# Check for the Running status of Ceph Pods
run_cmd(f"oc wait --for condition=ready pod "
f"-l app=rook-ceph-agent "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s")
assert pod.wait_for_resource(condition='Running',
selector='app=rook-ceph-mon',
resource_count=3,
timeout=600)
assert pod.wait_for_resource(condition='Running',
selector='app=rook-ceph-mgr',
timeout=600)
assert pod.wait_for_resource(condition='Running',
selector='app=rook-ceph-osd',
resource_count=3,
timeout=600)
create_oc_resource('toolbox.yaml', self.cluster_path, _templating,
config.ENV_DATA)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
create_oc_resource('storage-manifest.yaml', self.cluster_path,
_templating, config.ENV_DATA)
create_oc_resource("service-monitor.yaml", self.cluster_path,
_templating, config.ENV_DATA)
create_oc_resource("prometheus-rules.yaml", self.cluster_path,
_templating, config.ENV_DATA)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
# Create MDS pods for CephFileSystem
fs_data = templating.load_yaml_to_dict(constants.CEPHFILESYSTEM_YAML)
fs_data['metadata']['namespace'] = config.ENV_DATA['cluster_namespace']
ceph_obj = OCS(**fs_data)
ceph_obj.create()
assert pod.wait_for_resource(condition=constants.STATUS_RUNNING,
selector='app=rook-ceph-mds',
resource_count=2,
timeout=600)
# Check for CephFilesystem creation in ocp
cfs_data = cfs.get()
cfs_name = cfs_data['items'][0]['metadata']['name']
if helpers.validate_cephfilesystem(cfs_name):
logger.info(f"MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
logger.error(f"MDS deployment Failed! Please check logs!")
# Verify health of ceph cluster
# TODO: move destroy cluster logic to new CLI usage pattern?
logger.info("Done creating rook resources, waiting for HEALTH_OK")
assert ceph_health_check(
namespace=config.ENV_DATA['cluster_namespace'])
# patch gp2 (EBS) storage class as 'non-default'
logger.info("Patch gp2 storageclass as non-default")
patch = " '{\"metadata\": {\"annotations\":{\"storageclass.kubernetes.io/is-default-class\":\"false\"}}}' "
run_cmd(f"oc patch storageclass gp2 "
f"-p {patch} "
f"--request-timeout=120s")
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
_templating = templating.Templating()
ceph_cluster = ocp.OCP(
kind='CephCluster', namespace=config.ENV_DATA['cluster_namespace']
)
try:
ceph_cluster.get().get('items')[0]
logger.warning("OCS cluster already exists")
return
except (IndexError, CommandFailed):
logger.info("Running OCS basic installation")
create_oc_resource(
'common.yaml', self.cluster_path, _templating, config.ENV_DATA
)
run_cmd(
f'oc label namespace {config.ENV_DATA["cluster_namespace"]} '
f'"openshift.io/cluster-monitoring=true"'
)
run_cmd(
f"oc policy add-role-to-user view "
f"system:serviceaccount:openshift-monitoring:prometheus-k8s "
f"-n {config.ENV_DATA['cluster_namespace']}"
)
# HACK: If you would like to drop this hack, make sure that you also
# updated docs and write appropriate unit/integration tests for config
# processing.
if config.ENV_DATA.get('monitoring_enabled') in ("true", "True", True):
# RBAC rules for monitoring, based on documentation change in rook:
# https://github.com/rook/rook/commit/1b6fe840f6ae7372a9675ba727ecc65326708aa8
# HACK: This should be dropped when OCS is managed by OLM
apply_oc_resource(
'rbac.yaml',
self.cluster_path,
_templating,
config.ENV_DATA,
template_dir="monitoring"
)
# Increased to 15 seconds as 10 is not enough
# TODO: do the sampler function and check if resource exist
wait_time = 15
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
create_oc_resource(
'operator-openshift.yaml', self.cluster_path,
_templating, config.ENV_DATA
)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
run_cmd(
f"oc wait --for condition=ready pod "
f"-l app=rook-ceph-operator "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s"
)
run_cmd(
f"oc wait --for condition=ready pod "
f"-l app=rook-discover "
f"-n {config.ENV_DATA['cluster_namespace']} "
f"--timeout=120s"
)
create_oc_resource(
'cluster.yaml', self.cluster_path, _templating, config.ENV_DATA
)
pod = ocp.OCP(
kind=constants.POD, namespace=config.ENV_DATA['cluster_namespace']
)
cfs = ocp.OCP(
kind=constants.CEPHFILESYSTEM,
namespace=config.ENV_DATA['cluster_namespace']
)
# Check for Ceph pods
assert pod.wait_for_resource(
condition='Running', selector='app=rook-ceph-mon',
resource_count=3, timeout=600
)
assert pod.wait_for_resource(
condition='Running', selector='app=rook-ceph-mgr',
timeout=600
)
assert pod.wait_for_resource(
condition='Running', selector='app=rook-ceph-osd',
resource_count=3, timeout=600
)
create_oc_resource(
'toolbox.yaml', self.cluster_path, _templating, config.ENV_DATA
)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
# HACK: This should be dropped (including service-monitor.yaml and
# prometheus-rules.yaml files) when OCS is managed by OLM
if config.ENV_DATA.get('monitoring_enabled') not in ("true", "True", True):
# HACK: skip creation of rook-ceph-mgr service monitor when monitoring
# is enabled (if this were not skipped, the step would fail because
# rook would create the service monitor at this point already)
create_oc_resource(
"service-monitor.yaml", self.cluster_path, _templating,
config.ENV_DATA
)
# HACK: skip creation of prometheus-rules, rook-ceph is concerned
# with it's setup now, based on clarification from Umanga
# Chapagain
create_oc_resource(
"prometheus-rules.yaml", self.cluster_path, _templating,
config.ENV_DATA
)
logger.info(f"Waiting {wait_time} seconds...")
time.sleep(wait_time)
# Create MDS pods for CephFileSystem
fs_data = templating.load_yaml_to_dict(constants.CEPHFILESYSTEM_YAML)
fs_data['metadata']['namespace'] = config.ENV_DATA['cluster_namespace']
ceph_obj = OCS(**fs_data)
ceph_obj.create()
assert pod.wait_for_resource(
condition=constants.STATUS_RUNNING, selector='app=rook-ceph-mds',
resource_count=2, timeout=600
)
# Check for CephFilesystem creation in ocp
cfs_data = cfs.get()
cfs_name = cfs_data['items'][0]['metadata']['name']
if helpers.validate_cephfilesystem(cfs_name):
logger.info(f"MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
logger.error(
f"MDS deployment Failed! Please check logs!"
)
# Verify health of ceph cluster
# TODO: move destroy cluster logic to new CLI usage pattern?
logger.info("Done creating rook resources, waiting for HEALTH_OK")
assert ceph_health_check(
namespace=config.ENV_DATA['cluster_namespace']
)
# patch gp2 (EBS) storage class as 'non-default'
logger.info("Patch gp2 storageclass as non-default")
patch = " '{\"metadata\": {\"annotations\":{\"storageclass.kubernetes.io/is-default-class\":\"false\"}}}' "
run_cmd(
f"oc patch storageclass gp2 "
f"-p {patch} "
f"--request-timeout=120s"
)