def get_ceph_tools_pod():
"""
Get the Ceph tools pod
Returns:
Pod object: The Ceph tools pod object
"""
ocp_pod_obj = OCP(kind=constants.POD,
namespace=config.ENV_DATA['cluster_namespace'])
ct_pod_items = ocp_pod_obj.get(selector='app=rook-ceph-tools')['items']
if not ct_pod_items:
# setup ceph_toolbox pod if the cluster has been setup by some other CI
setup_ceph_toolbox()
ct_pod_items = ocp_pod_obj.get(selector='app=rook-ceph-tools')['items']
assert ct_pod_items, "No Ceph tools pod found"
# In the case of node failure, the CT pod will be recreated with the old
# one in status Terminated. Therefore, need to filter out the Terminated pod
running_ct_pods = list()
for pod in ct_pod_items:
if ocp_pod_obj.get_resource_status(
pod.get('metadata').get('name')) == constants.STATUS_RUNNING:
running_ct_pods.append(pod)
assert running_ct_pods, "No running Ceph tools pod found"
ceph_pod = Pod(**running_ct_pods[0])
return ceph_pod
def deploy_with_external_mode(self):
"""
This function handles the deployment of OCS on
external/indpendent RHCS cluster
"""
live_deployment = config.DEPLOYMENT.get("live_deployment")
logger.info("Deploying OCS with external mode RHCS")
ui_deployment = config.DEPLOYMENT.get("ui_deployment")
if not ui_deployment:
logger.info("Creating namespace and operator group.")
run_cmd(f"oc create -f {constants.OLM_YAML}")
if not live_deployment:
self.create_ocs_operator_source()
self.subscribe_ocs()
operator_selector = get_selector_for_ocs_operator()
subscription_plan_approval = config.DEPLOYMENT.get(
"subscription_plan_approval")
package_manifest = PackageManifest(
resource_name=defaults.OCS_OPERATOR_NAME,
selector=operator_selector,
subscription_plan_approval=subscription_plan_approval,
)
package_manifest.wait_for_resource(timeout=300)
channel = config.DEPLOYMENT.get("ocs_csv_channel")
csv_name = package_manifest.get_current_csv(channel=channel)
csv = CSV(resource_name=csv_name, namespace=self.namespace)
csv.wait_for_phase("Succeeded", timeout=720)
# Create secret for external cluster
secret_data = templating.load_yaml(
constants.EXTERNAL_CLUSTER_SECRET_YAML)
external_cluster_details = config.EXTERNAL_MODE.get(
"external_cluster_details", "")
if not external_cluster_details:
raise ExternalClusterDetailsException(
"No external cluster data found")
secret_data["data"][
"external_cluster_details"] = external_cluster_details
secret_data_yaml = tempfile.NamedTemporaryFile(
mode="w+", prefix="external_cluster_secret", delete=False)
templating.dump_data_to_temp_yaml(secret_data, secret_data_yaml.name)
logger.info("Creating external cluster secret")
run_cmd(f"oc create -f {secret_data_yaml.name}")
cluster_data = templating.load_yaml(
constants.EXTERNAL_STORAGE_CLUSTER_YAML)
cluster_data["metadata"]["name"] = config.ENV_DATA[
"storage_cluster_name"]
cluster_data_yaml = tempfile.NamedTemporaryFile(
mode="w+", prefix="external_cluster_storage", delete=False)
templating.dump_data_to_temp_yaml(cluster_data, cluster_data_yaml.name)
run_cmd(f"oc create -f {cluster_data_yaml.name}", timeout=2400)
self.external_post_deploy_validation()
setup_ceph_toolbox()
def deploy_with_external_mode(self):
"""
This function handles the deployment of OCS on
external/indpendent RHCS cluster
"""
live_deployment = config.DEPLOYMENT.get("live_deployment")
logger.info("Deploying OCS with external mode RHCS")
ui_deployment = config.DEPLOYMENT.get("ui_deployment")
if not ui_deployment:
logger.info("Creating namespace and operator group.")
run_cmd(f"oc create -f {constants.OLM_YAML}")
if not live_deployment:
create_catalog_source()
self.subscribe_ocs()
operator_selector = get_selector_for_ocs_operator()
subscription_plan_approval = config.DEPLOYMENT.get("subscription_plan_approval")
ocs_version = version.get_semantic_ocs_version_from_config()
if ocs_version >= version.VERSION_4_9:
ocs_operator_names = [
defaults.ODF_OPERATOR_NAME,
defaults.OCS_OPERATOR_NAME,
]
else:
ocs_operator_names = [defaults.OCS_OPERATOR_NAME]
channel = config.DEPLOYMENT.get("ocs_csv_channel")
for ocs_operator_name in ocs_operator_names:
package_manifest = PackageManifest(
resource_name=ocs_operator_name,
selector=operator_selector,
subscription_plan_approval=subscription_plan_approval,
)
package_manifest.wait_for_resource(timeout=300)
csv_name = package_manifest.get_current_csv(channel=channel)
csv = CSV(resource_name=csv_name, namespace=self.namespace)
csv.wait_for_phase("Succeeded", timeout=720)
# Set rook log level
self.set_rook_log_level()
# Create secret for external cluster
create_external_secret()
cluster_data = templating.load_yaml(constants.EXTERNAL_STORAGE_CLUSTER_YAML)
cluster_data["metadata"]["name"] = config.ENV_DATA["storage_cluster_name"]
cluster_data_yaml = tempfile.NamedTemporaryFile(
mode="w+", prefix="external_cluster_storage", delete=False
)
templating.dump_data_to_temp_yaml(cluster_data, cluster_data_yaml.name)
run_cmd(f"oc create -f {cluster_data_yaml.name}", timeout=2400)
self.external_post_deploy_validation()
setup_ceph_toolbox()
def get_images_post_upgrade(self, channel, pre_upgrade_images,
upgrade_version):
"""
Checks if all images of OCS cluster upgraded,
and return list of all images if upgrade success
Args:
channel: (str): OCS subscription channel
pre_upgrade_images: (dict): Contains all OCS cluster images
upgrade_version: (str): version to be upgraded
Returns:
set: Contains full path of OCS cluster old images
"""
operator_selector = get_selector_for_ocs_operator()
package_manifest = PackageManifest(
resource_name=OCS_OPERATOR_NAME,
selector=operator_selector,
subscription_plan_approval=self.subscription_plan_approval,
)
csv_name_post_upgrade = package_manifest.get_current_csv(channel)
csv_post_upgrade = CSV(resource_name=csv_name_post_upgrade,
namespace=self.namespace)
log.info(
f"Waiting for CSV {csv_name_post_upgrade} to be in succeeded state"
)
# Workaround for patching missing ceph-rook-tools pod after upgrade
if self.version_before_upgrade == "4.2" and upgrade_version == "4.3":
log.info("Force creating Ceph toolbox after upgrade 4.2 -> 4.3")
setup_ceph_toolbox(force_setup=True)
# End of workaround
if config.DEPLOYMENT.get("external_mode") or config.ENV_DATA.get(
"mcg_only_deployment"):
timeout = 200
else:
timeout = 200 * get_osd_count()
csv_post_upgrade.wait_for_phase("Succeeded", timeout=timeout)
post_upgrade_images = get_images(csv_post_upgrade.get())
old_images, _, _ = get_upgrade_image_info(pre_upgrade_images,
post_upgrade_images)
return old_images
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
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")
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()
# validate PDB creation of MON, MDS, OSD pods
validate_pdb_creation()
# 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)
# 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("MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
logger.error("MDS deployment Failed! Please check logs!")
# Change monitoring backend to OCS
if config.ENV_DATA.get('monitoring_enabled') and config.ENV_DATA.get(
'persistent-monitoring'):
sc = helpers.default_storage_class(
interface_type=constants.CEPHBLOCKPOOL)
# Get the list of monitoring pods
pods_list = get_all_pods(
namespace=defaults.OCS_MONITORING_NAMESPACE,
selector=['prometheus', 'alertmanager'])
# Create configmap cluster-monitoring-config and reconfigure
# storage class and telemeter server (if the url is specified in a
# config file)
create_configmap_cluster_monitoring_pod(
sc_name=sc.name,
telemeter_server_url=config.ENV_DATA.get(
"telemeter_server_url"))
# Take some time to respin the pod
waiting_time = 45
logger.info(f"Waiting {waiting_time} seconds...")
time.sleep(waiting_time)
# Validate the pods are respinned and in running state
retry((CommandFailed, ResourceWrongStatusException),
tries=3,
delay=15)(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
retry((CommandFailed, AssertionError), tries=3, delay=15)(
validate_pvc_are_mounted_on_monitoring_pods)(pods_list)
elif config.ENV_DATA.get('monitoring_enabled') and config.ENV_DATA.get(
"telemeter_server_url"):
# Create configmap cluster-monitoring-config to reconfigure
# telemeter server url when 'persistent-monitoring' is False
create_configmap_cluster_monitoring_pod(
telemeter_server_url=config.ENV_DATA["telemeter_server_url"])
# 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")
try:
ceph_health_check(namespace=self.namespace, tries=30, delay=10)
except CephHealthException as ex:
err = str(ex)
logger.warning(f"Ceph health check failed with {err}")
if "clock skew detected" in err:
logger.info(f"Changing NTP on compute nodes to"
f" {constants.RH_NTP_CLOCK}")
update_ntp_compute_nodes()
assert ceph_health_check(namespace=self.namespace,
tries=60,
delay=10)
# patch gp2/thin storage class as 'non-default'
self.patch_default_sc_to_non_default()
if check_nodes_specs(min_cpu=constants.MIN_NODE_CPU,
min_memory=constants.MIN_NODE_MEMORY):
logger.info("The cluster specs meet the minimum requirements and "
"therefore, NooBaa auto scale will be enabled")
min_nb_eps = config.DEPLOYMENT.get('min_noobaa_endpoints')
max_nb_eps = config.DEPLOYMENT.get('max_noobaa_endpoints')
change_noobaa_endpoints_count(min_nb_eps=min_nb_eps,
max_nb_eps=max_nb_eps)
else:
logger.warning(
"The cluster specs do not meet the minimum requirements"
" and therefore, NooBaa auto scale will remain disabled")
change_noobaa_endpoints_count(min_nb_eps=1, max_nb_eps=1)
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()
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
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")
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()
# validate PDB creation of MON, MDS, OSD pods
validate_pdb_creation()
# 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)
# Workaround for https://bugzilla.redhat.com/show_bug.cgi?id=1847098
if config.DEPLOYMENT.get('local_storage'):
tools_pod = run_cmd(
f"oc -n {self.namespace} get pod -l 'app=rook-ceph-tools' "
f"-o jsonpath='{{.items[0].metadata.name}}'")
pgs_to_autoscale = [
'ocs-storagecluster-cephblockpool',
'ocs-storagecluster-cephfilesystem-data0'
]
for pg in pgs_to_autoscale:
run_cmd(f"oc -n {self.namespace} exec {tools_pod} -- "
f"ceph osd pool set {pg} pg_autoscale_mode on")
# 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("MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
logger.error("MDS deployment Failed! Please check logs!")
# Change monitoring backend to OCS
if config.ENV_DATA.get('monitoring_enabled') and config.ENV_DATA.get(
'persistent-monitoring'):
sc = helpers.default_storage_class(
interface_type=constants.CEPHBLOCKPOOL)
# Get the list of monitoring pods
pods_list = get_all_pods(
namespace=defaults.OCS_MONITORING_NAMESPACE,
selector=['prometheus', 'alertmanager'])
# Create configmap cluster-monitoring-config and reconfigure
# storage class and telemeter server (if the url is specified in a
# config file)
create_configmap_cluster_monitoring_pod(
sc_name=sc.name,
telemeter_server_url=config.ENV_DATA.get(
"telemeter_server_url"))
# Take some time to respin the pod
waiting_time = 45
logger.info(f"Waiting {waiting_time} seconds...")
time.sleep(waiting_time)
# Validate the pods are respinned and in running state
retry((CommandFailed, ResourceWrongStatusException),
tries=3,
delay=15)(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
retry((CommandFailed, AssertionError), tries=3, delay=15)(
validate_pvc_are_mounted_on_monitoring_pods)(pods_list)
elif config.ENV_DATA.get('monitoring_enabled') and config.ENV_DATA.get(
"telemeter_server_url"):
# Create configmap cluster-monitoring-config to reconfigure
# telemeter server url when 'persistent-monitoring' is False
create_configmap_cluster_monitoring_pod(
telemeter_server_url=config.ENV_DATA["telemeter_server_url"])
# 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()
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
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 config.DEPLOYMENT["external_mode"]:
logger.info("Deploying OCS on external mode RHCS")
return self.deploy_with_external_mode()
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()
# validate PDB creation of MON, MDS, OSD pods
validate_pdb_creation()
# 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,
)
# 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("MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
logger.error("MDS deployment Failed! Please check logs!")
# Change monitoring backend to OCS
if config.ENV_DATA.get("monitoring_enabled") and config.ENV_DATA.get(
"persistent-monitoring"):
sc = helpers.default_storage_class(
interface_type=constants.CEPHBLOCKPOOL)
# Get the list of monitoring pods
pods_list = get_all_pods(
namespace=defaults.OCS_MONITORING_NAMESPACE,
selector=["prometheus", "alertmanager"],
)
# Create configmap cluster-monitoring-config and reconfigure
# storage class and telemeter server (if the url is specified in a
# config file)
create_configmap_cluster_monitoring_pod(
sc_name=sc.name,
telemeter_server_url=config.ENV_DATA.get(
"telemeter_server_url"),
)
# Take some time to respin the pod
waiting_time = 45
logger.info(f"Waiting {waiting_time} seconds...")
time.sleep(waiting_time)
# Validate the pods are respinned and in running state
retry((CommandFailed, ResourceWrongStatusException),
tries=3,
delay=15)(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
retry((CommandFailed, AssertionError), tries=3, delay=15)(
validate_pvc_are_mounted_on_monitoring_pods)(pods_list)
elif config.ENV_DATA.get("monitoring_enabled") and config.ENV_DATA.get(
"telemeter_server_url"):
# Create configmap cluster-monitoring-config to reconfigure
# telemeter server url when 'persistent-monitoring' is False
create_configmap_cluster_monitoring_pod(
telemeter_server_url=config.ENV_DATA["telemeter_server_url"])
# 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")
try:
ceph_health_check(namespace=self.namespace, tries=30, delay=10)
except CephHealthException as ex:
err = str(ex)
logger.warning(f"Ceph health check failed with {err}")
if "clock skew detected" in err:
logger.info(f"Changing NTP on compute nodes to"
f" {constants.RH_NTP_CLOCK}")
if self.platform == constants.VSPHERE_PLATFORM:
update_ntp_compute_nodes()
assert ceph_health_check(namespace=self.namespace,
tries=60,
delay=10)
# patch gp2/thin storage class as 'non-default'
self.patch_default_sc_to_non_default()
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
set_registry_to_managed_state()
image = None
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")
# disconnected installation?
load_cluster_info()
if config.DEPLOYMENT.get("disconnected"):
image = prepare_disconnected_ocs_deployment()
if config.DEPLOYMENT["external_mode"]:
self.deploy_with_external_mode()
else:
self.deploy_ocs_via_operator(image)
pod = ocp.OCP(kind=constants.POD, namespace=self.namespace)
cfs = ocp.OCP(kind=constants.CEPHFILESYSTEM,
namespace=self.namespace)
# Check for Ceph pods
mon_pod_timeout = (900 if self.platform
== constants.IBMCLOUD_PLATFORM else 600)
assert pod.wait_for_resource(
condition="Running",
selector="app=rook-ceph-mon",
resource_count=3,
timeout=mon_pod_timeout,
)
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()
# validate PDB creation of MON, MDS, OSD pods
validate_pdb_creation()
# 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 config.COMPONENTS["disable_cephfs"]:
# 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("MDS deployment is successful!")
defaults.CEPHFILESYSTEM_NAME = cfs_name
else:
logger.error("MDS deployment Failed! Please check logs!")
# Change monitoring backend to OCS
if config.ENV_DATA.get("monitoring_enabled") and config.ENV_DATA.get(
"persistent-monitoring"):
setup_persistent_monitoring()
elif config.ENV_DATA.get("monitoring_enabled") and config.ENV_DATA.get(
"telemeter_server_url"):
# Create configmap cluster-monitoring-config to reconfigure
# telemeter server url when 'persistent-monitoring' is False
create_configmap_cluster_monitoring_pod(
telemeter_server_url=config.ENV_DATA["telemeter_server_url"])
if not config.COMPONENTS["disable_cephfs"]:
# Change registry backend to OCS CEPHFS RWX PVC
registry.change_registry_backend_to_ocs()
# Verify health of ceph cluster
logger.info("Done creating rook resources, waiting for HEALTH_OK")
try:
ceph_health_check(namespace=self.namespace, tries=30, delay=10)
except CephHealthException as ex:
err = str(ex)
logger.warning(f"Ceph health check failed with {err}")
if "clock skew detected" in err:
logger.info(f"Changing NTP on compute nodes to"
f" {constants.RH_NTP_CLOCK}")
if self.platform == constants.VSPHERE_PLATFORM:
update_ntp_compute_nodes()
assert ceph_health_check(namespace=self.namespace,
tries=60,
delay=10)
# patch gp2/thin storage class as 'non-default'
self.patch_default_sc_to_non_default()
def test_upgrade():
ceph_cluster = CephCluster()
with CephHealthMonitor(ceph_cluster):
namespace = config.ENV_DATA['cluster_namespace']
version_before_upgrade = config.ENV_DATA.get("ocs_version")
upgrade_version = config.UPGRADE.get("upgrade_ocs_version",
version_before_upgrade)
ocs_registry_image = config.UPGRADE.get('upgrade_ocs_registry_image')
if ocs_registry_image:
upgrade_version = get_ocs_version_from_image(ocs_registry_image)
parsed_version_before_upgrade = parse_version(version_before_upgrade)
parsed_upgrade_version = parse_version(upgrade_version)
assert parsed_upgrade_version >= parsed_version_before_upgrade, (
f"Version you would like to upgrade to: {upgrade_version} "
f"is not higher or equal to the version you currently running: "
f"{version_before_upgrade}")
operator_selector = get_selector_for_ocs_operator()
package_manifest = PackageManifest(
resource_name=OCS_OPERATOR_NAME,
selector=operator_selector,
)
channel = config.DEPLOYMENT.get('ocs_csv_channel')
csv_name_pre_upgrade = package_manifest.get_current_csv(channel)
log.info(f"CSV name before upgrade is: {csv_name_pre_upgrade}")
csv_pre_upgrade = CSV(resource_name=csv_name_pre_upgrade,
namespace=namespace)
pre_upgrade_images = get_images(csv_pre_upgrade.get())
version_change = parsed_upgrade_version > parsed_version_before_upgrade
if version_change:
version_config_file = os.path.join(constants.CONF_DIR,
'ocs_version',
f'ocs-{upgrade_version}.yaml')
load_config_file(version_config_file)
ocs_catalog = CatalogSource(
resource_name=constants.OPERATOR_CATALOG_SOURCE_NAME,
namespace=constants.MARKETPLACE_NAMESPACE,
)
upgrade_in_current_source = config.UPGRADE.get(
'upgrade_in_current_source', False)
if not upgrade_in_current_source:
if not ocs_catalog.is_exist() and not upgrade_in_current_source:
log.info("OCS catalog source doesn't exist. Creating new one.")
create_catalog_source(ocs_registry_image, ignore_upgrade=True)
image_url = ocs_catalog.get_image_url()
image_tag = ocs_catalog.get_image_name()
log.info(f"Current image is: {image_url}, tag: {image_tag}")
if ocs_registry_image:
image_url, new_image_tag = ocs_registry_image.split(':')
elif (config.UPGRADE.get('upgrade_to_latest', True)
or version_change):
new_image_tag = get_latest_ds_olm_tag()
else:
new_image_tag = get_next_version_available_for_upgrade(
image_tag)
cs_data = deepcopy(ocs_catalog.data)
image_for_upgrade = ':'.join([image_url, new_image_tag])
log.info(f"Image: {image_for_upgrade} will be used for upgrade.")
cs_data['spec']['image'] = image_for_upgrade
with NamedTemporaryFile() as cs_yaml:
dump_data_to_temp_yaml(cs_data, cs_yaml.name)
ocs_catalog.apply(cs_yaml.name)
# Wait for the new package manifest for upgrade.
operator_selector = get_selector_for_ocs_operator()
package_manifest = PackageManifest(
resource_name=OCS_OPERATOR_NAME,
selector=operator_selector,
)
package_manifest.wait_for_resource()
channel = config.DEPLOYMENT.get('ocs_csv_channel')
if not channel:
channel = package_manifest.get_default_channel()
# update subscription
subscription = OCP(
resource_name=constants.OCS_SUBSCRIPTION,
kind='subscription',
namespace=config.ENV_DATA['cluster_namespace'],
)
current_ocs_source = subscription.data['spec']['source']
log.info(f"Current OCS subscription source: {current_ocs_source}")
ocs_source = current_ocs_source if upgrade_in_current_source else (
constants.OPERATOR_CATALOG_SOURCE_NAME)
patch_subscription_cmd = (
f'oc patch subscription {constants.OCS_SUBSCRIPTION} '
f'-n {namespace} --type merge -p \'{{"spec":{{"channel": '
f'"{channel}", "source": "{ocs_source}"}}}}\'')
run_cmd(patch_subscription_cmd)
subscription_plan_approval = config.DEPLOYMENT.get(
'subscription_plan_approval')
if subscription_plan_approval == 'Manual':
wait_for_install_plan_and_approve(namespace)
attempts = 145
for attempt in range(1, attempts + 1):
log.info(f"Attempt {attempt}/{attempts} to check CSV upgraded.")
csv_name_post_upgrade = package_manifest.get_current_csv(channel)
if csv_name_post_upgrade == csv_name_pre_upgrade:
log.info(f"CSV is still: {csv_name_post_upgrade}")
sleep(5)
else:
log.info(f"CSV now upgraded to: {csv_name_post_upgrade}")
break
if attempts == attempt:
raise TimeoutException("No new CSV found after upgrade!")
csv_post_upgrade = CSV(resource_name=csv_name_post_upgrade,
namespace=namespace)
log.info(
f"Waiting for CSV {csv_name_post_upgrade} to be in succeeded state"
)
if version_before_upgrade == '4.2' and upgrade_version == '4.3':
log.info("Force creating Ceph toolbox after upgrade 4.2 -> 4.3")
setup_ceph_toolbox(force_setup=True)
osd_count = get_osd_count()
csv_post_upgrade.wait_for_phase("Succeeded", timeout=200 * osd_count)
post_upgrade_images = get_images(csv_post_upgrade.get())
old_images, _, _ = get_upgrade_image_info(pre_upgrade_images,
post_upgrade_images)
verify_image_versions(old_images, parsed_upgrade_version)
ocs_install_verification(
timeout=600,
skip_osd_distribution_check=True,
ocs_registry_image=ocs_registry_image,
post_upgrade_verification=True,
)
def test_upgrade():
ceph_cluster = CephCluster()
with CephHealthMonitor(ceph_cluster):
namespace = config.ENV_DATA['cluster_namespace']
ocs_catalog = CatalogSource(
resource_name=constants.OPERATOR_CATALOG_SOURCE_NAME,
namespace=constants.MARKETPLACE_NAMESPACE,
)
version_before_upgrade = config.ENV_DATA.get("ocs_version")
upgrade_version = config.UPGRADE.get("upgrade_ocs_version",
version_before_upgrade)
parsed_version_before_upgrade = parse_version(version_before_upgrade)
parsed_upgrade_version = parse_version(upgrade_version)
assert parsed_upgrade_version >= parsed_version_before_upgrade, (
f"Version you would like to upgrade to: {upgrade_version} "
f"is not higher or equal to the version you currently running: "
f"{version_before_upgrade}")
version_change = parsed_upgrade_version > parsed_version_before_upgrade
if version_change:
version_config_file = os.path.join(constants.CONF_DIR,
'ocs_version',
f'ocs-{upgrade_version}.yaml')
assert os.path.exists(version_config_file), (
f"OCS version config file {version_config_file} doesn't exist!"
)
with open(os.path.abspath(
os.path.expanduser(version_config_file))) as file_stream:
custom_config_data = yaml.safe_load(file_stream)
config.update(custom_config_data)
image_url = ocs_catalog.get_image_url()
image_tag = ocs_catalog.get_image_name()
log.info(f"Current image is: {image_url}, tag: {image_tag}")
ocs_registry_image = config.UPGRADE.get('upgrade_ocs_registry_image')
if ocs_registry_image:
image_url, new_image_tag = ocs_registry_image.split(':')
elif config.UPGRADE.get('upgrade_to_latest', True) or version_change:
new_image_tag = get_latest_ds_olm_tag()
else:
new_image_tag = get_next_version_available_for_upgrade(image_tag)
cs_data = deepcopy(ocs_catalog.data)
image_for_upgrade = ':'.join([image_url, new_image_tag])
log.info(f"Image: {image_for_upgrade} will be used for upgrade.")
cs_data['spec']['image'] = image_for_upgrade
operator_selector = get_selector_for_ocs_operator()
package_manifest = PackageManifest(
resource_name=OCS_OPERATOR_NAME,
selector=operator_selector,
)
csv_name_pre_upgrade = package_manifest.get_current_csv()
log.info(f"CSV name before upgrade is: {csv_name_pre_upgrade}")
csv_pre_upgrade = CSV(resource_name=csv_name_pre_upgrade,
namespace=namespace)
pre_upgrade_images = get_images(csv_pre_upgrade.get())
with NamedTemporaryFile() as cs_yaml:
dump_data_to_temp_yaml(cs_data, cs_yaml.name)
ocs_catalog.apply(cs_yaml.name)
# Wait for package manifest is ready
package_manifest.wait_for_resource()
subscription_plan_approval = config.DEPLOYMENT.get(
'subscription_plan_approval')
if subscription_plan_approval == 'Manual':
wait_for_install_plan_and_approve(namespace)
attempts = 145
for attempt in range(1, attempts):
if attempts == attempt:
raise TimeoutException("No new CSV found after upgrade!")
log.info(f"Attempt {attempt}/{attempts} to check CSV upgraded.")
package_manifest.reload_data()
csv_name_post_upgrade = package_manifest.get_current_csv()
if csv_name_post_upgrade == csv_name_pre_upgrade:
log.info(f"CSV is still: {csv_name_post_upgrade}")
sleep(5)
else:
log.info(f"CSV now upgraded to: {csv_name_post_upgrade}")
break
csv_post_upgrade = CSV(resource_name=csv_name_post_upgrade,
namespace=namespace)
log.info(
f"Waiting for CSV {csv_name_post_upgrade} to be in succeeded state"
)
if version_before_upgrade == '4.2' and upgrade_version == '4.3':
log.info("Force creating Ceph toolbox after upgrade 4.2 -> 4.3")
setup_ceph_toolbox(force_setup=True)
csv_post_upgrade.wait_for_phase("Succeeded", timeout=600)
post_upgrade_images = get_images(csv_post_upgrade.get())
old_images, _, _ = get_upgrade_image_info(pre_upgrade_images,
post_upgrade_images)
verify_image_versions(old_images, parsed_upgrade_version)
ocs_install_verification(timeout=600, skip_osd_distribution_check=True)
def deploy_ocs(self):
"""
Handle OCS deployment, since OCS deployment steps are common to any
platform, implementing OCS deployment here in base class.
"""
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")
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
)
# 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!"
)
# Change monitoring backend to OCS
if config.ENV_DATA.get('monitoring_enabled') and config.ENV_DATA.get('persistent-monitoring'):
sc = helpers.default_storage_class(interface_type=constants.CEPHBLOCKPOOL)
# Get the list of monitoring pods
pods_list = get_all_pods(
namespace=defaults.OCS_MONITORING_NAMESPACE,
selector=['prometheus', 'alertmanager']
)
# Create configmap cluster-monitoring-config and reconfigure
# storage class and telemeter server (if the url is specified in a
# config file)
create_configmap_cluster_monitoring_pod(
sc_name=sc.name,
telemeter_server_url=config.ENV_DATA.get("telemeter_server_url"))
# Take some time to respin the pod
waiting_time = 45
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)
elif config.ENV_DATA.get('monitoring_enabled') and config.ENV_DATA.get("telemeter_server_url"):
# Create configmap cluster-monitoring-config to reconfigure
# telemeter server url when 'persistent-monitoring' is False
create_configmap_cluster_monitoring_pod(
telemeter_server_url=config.ENV_DATA["telemeter_server_url"])
# 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()
