def test_automated_recovery_from_failed_nodes_IPI_proactive(
self, pvc_factory, pod_factory):
"""
Knip-678 Automated recovery from failed nodes
Proactive case - IPI
"""
# Get the osd associated node name
osd_pods_obj = pod.get_osd_pods()
osd_node_name = pod.get_pod_node(random.choice(osd_pods_obj)).name
log.info(f"Selected OSD is {osd_node_name}")
# Get the machine name using the node name
machine_name = machine.get_machine_from_node_name(osd_node_name)
log.info(f"{osd_node_name} associated machine is {machine_name}")
# Get the machineset name using machine name
machineset_name = machine.get_machineset_from_machine_name(
machine_name)
log.info(f"{osd_node_name} associated machineset is {machineset_name}")
# Add a new node and label it
add_new_node_and_label_it(machineset_name)
# Delete the machine
machine.delete_machine(machine_name)
log.info(f"Successfully deleted machine {machine_name}")
# Check the pods should be in running state
all_pod_obj = pod.get_all_pods(wait=True)
for pod_obj in all_pod_obj:
if '-1-deploy' and 'ocs-deviceset' not in pod_obj.name:
try:
wait_for_resource_state(resource=pod_obj,
state=constants.STATUS_RUNNING,
timeout=200)
# 'rook-ceph-crashcollector' on the failed node stucks at pending
# state. BZ 1810014 tracks it.
# Ignoring 'rook-ceph-crashcollector' pod health check as WA and
# deleting its deployment so that the pod disappears
# Will revert this WA once the BZ is fixed
except ResourceWrongStatusException:
if 'rook-ceph-crashcollector' in pod_obj.name:
ocp_obj = ocp.OCP()
name = pod_obj.name[:-17]
command = f"delete deployment {name}"
ocp_obj.exec_oc_cmd(command=command)
log.info(f"Deleted deployment for pod {pod_obj.name}")
# Check basic cluster functionality by creating resources
# (pools, storageclasses, PVCs, pods - both CephFS and RBD),
# run IO and delete the resources
self.sanity_helpers.create_resources(pvc_factory, pod_factory)
self.sanity_helpers.delete_resources()
# Perform cluster and Ceph health checks
self.sanity_helpers.health_check()
def test_automated_recovery_from_failed_nodes_IPI_proactive(
self, pvc_factory, pod_factory):
"""
Knip-678 Automated recovery from failed nodes
Proactive case - IPI
"""
# Get the osd associated node name
osd_pods_obj = pod.get_osd_pods()
osd_node_name = pod.get_pod_node(random.choice(osd_pods_obj)).name
log.info(f"Selected OSD is {osd_node_name}")
# Get the machine name using the node name
machine_name = machine.get_machine_from_node_name(osd_node_name)
log.info(f"{osd_node_name} associated machine is {machine_name}")
# Get the machineset name using machine name
machineset_name = machine.get_machineset_from_machine_name(
machine_name)
log.info(f"{osd_node_name} associated machineset is {machineset_name}")
# Add a new node and label it
add_new_node_and_label_it(machineset_name)
# Delete the machine
machine.delete_machine(machine_name)
log.info(f"Successfully deleted machine {machine_name}")
# Check the pods should be in running state
all_pod_obj = pod.get_all_pods(wait=True)
for pod_obj in all_pod_obj:
wait_for_resource_state(resource=pod_obj,
state=constants.STATUS_RUNNING,
timeout=200)
# Check basic cluster functionality by creating resources
# (pools, storageclasses, PVCs, pods - both CephFS and RBD),
# run IO and delete the resources
self.sanity_helpers.create_resources(pvc_factory, pod_factory)
self.sanity_helpers.delete_resources()
# Perform cluster and Ceph health checks
self.sanity_helpers.health_check()
def test_automated_recovery_from_failed_nodes_IPI_proactive(
self, interface, pvc_factory, pod_factory, dc_pod_factory):
"""
Knip-678 Automated recovery from failed nodes
Proactive case - IPI
"""
# Get OSD running nodes
osd_running_nodes = get_osd_running_nodes()
log.info(f"OSDs are running on nodes {osd_running_nodes}")
# Label osd nodes with fedora app
label_worker_node(osd_running_nodes,
label_key='dc',
label_value='fedora')
# Create DC app pods
log.info("Creating DC based app pods")
interface = (constants.CEPHBLOCKPOOL
if interface == 'rbd' else constants.CEPHFILESYSTEM)
dc_pod_obj = []
for i in range(2):
dc_pod = dc_pod_factory(interface=interface,
node_selector={'dc': 'fedora'})
pod.run_io_in_bg(dc_pod, fedora_dc=True)
dc_pod_obj.append(dc_pod)
# Get app pods running nodes
dc_pod_node_name = get_app_pod_running_nodes(dc_pod_obj)
log.info(f"DC app pod running nodes are {dc_pod_node_name}")
# Get both osd and app pod running node
common_nodes = get_both_osd_and_app_pod_running_node(
osd_running_nodes, dc_pod_node_name)
msg = "Common OSD and app running node(s) NOT found"
assert (len(common_nodes) > 0), msg
log.info(f"Common OSD and app pod running nodes are {common_nodes}")
# Get the machine name using the node name
machine_name = machine.get_machine_from_node_name(common_nodes[0])
log.info(f"{common_nodes[0]} associated machine is {machine_name}")
# Get the machineset name using machine name
machineset_name = machine.get_machineset_from_machine_name(
machine_name)
log.info(
f"{common_nodes[0]} associated machineset is {machineset_name}")
# Add a new node and label it
add_new_node_and_label_it(machineset_name)
# Delete the machine
machine.delete_machine(machine_name)
log.info(f"Successfully deleted machine {machine_name}")
# DC app pods on the failed node will get automatically created on
# other running node. Waiting for all dc app pod to reach running
# state
pod.wait_for_dc_app_pods_to_reach_running_state(dc_pod_obj)
log.info("All the dc pods reached running state")
pod.wait_for_storage_pods()
# Check basic cluster functionality by creating resources
# (pools, storageclasses, PVCs, pods - both CephFS and RBD),
# run IO and delete the resources
self.sanity_helpers.create_resources(pvc_factory, pod_factory)
self.sanity_helpers.delete_resources()
# Perform cluster and Ceph health checks
self.sanity_helpers.health_check()
def test_node_replacement_reactive_aws_ipi(
self,
nodes,
pvc_factory,
pod_factory,
dc_pod_factory,
failure,
interface,
bucket_factory,
rgw_bucket_factory,
):
"""
Knip-894 Node replacement - AWS-IPI-Reactive
"""
# Get worker nodes
initial_nodes = get_worker_nodes()
# Get OSD running nodes
osd_running_nodes = get_osd_running_nodes()
log.info(f"OSDs are running on nodes {osd_running_nodes}")
# Label osd nodes with fedora app
label_worker_node(osd_running_nodes,
label_key="dc",
label_value="fedora")
# Create DC app pods
log.info("Creating DC based app pods")
if interface == "rbd":
interface = constants.CEPHBLOCKPOOL
elif interface == "cephfs":
interface = constants.CEPHFILESYSTEM
dc_pod_obj = []
for i in range(2):
dc_pod = dc_pod_factory(interface=interface,
node_selector={"dc": "fedora"})
pod.run_io_in_bg(dc_pod, fedora_dc=True)
dc_pod_obj.append(dc_pod)
# Get app pods running nodes
dc_pod_node_name = get_app_pod_running_nodes(dc_pod_obj)
log.info(f"DC app pod running nodes are {dc_pod_node_name}")
# Get both osd and app pod running node
common_nodes = get_both_osd_and_app_pod_running_node(
osd_running_nodes, dc_pod_node_name)
log.info(f"Both OSD and app pod is running on nodes {common_nodes}")
# Get the machine name using the node name
machine_name = machine.get_machine_from_node_name(common_nodes[0])
log.info(f"{common_nodes[0]} associated machine is {machine_name}")
# Get the machineset name using machine name
machineset_name = machine.get_machineset_from_machine_name(
machine_name)
log.info(
f"{common_nodes[0]} associated machineset is {machineset_name}")
# Get the failure node obj
failure_node_obj = get_node_objs(node_names=[common_nodes[0]])
# Induce failure on the selected failure node
log.info(f"Inducing failure on node {failure_node_obj[0].name}")
if failure == "power off":
# Power off AWS worker node instance
nodes.stop_nodes(failure_node_obj, wait=True)
log.info(
f"Successfully powered off node: {failure_node_obj[0].name}")
elif failure == "network failure":
# Induce Network failure
node_network_failure([failure_node_obj[0].name])
# Add annotation to the failed node
annotation = "machine.openshift.io/exclude-node-draining=''"
machine.add_annotation_to_machine(annotation=annotation,
machine_name=machine_name)
# Delete the machine
machine.delete_machine(machine_name)
log.info(f"Successfully deleted machine {machine_name}")
# Wait for the new machine to spin
log.info("Waiting for the new node to be in ready state")
machine.wait_for_new_node_to_be_ready(machineset_name)
# Get the node name of new spun node
nodes_after_new_spun_node = get_worker_nodes()
new_spun_node = list(
set(nodes_after_new_spun_node) - set(initial_nodes))
log.info(f"New spun node is {new_spun_node}")
# Label it
node_obj = ocp.OCP(kind="node")
node_obj.add_label(resource_name=new_spun_node[0],
label=constants.OPERATOR_NODE_LABEL)
log.info(
f"Successfully labeled {new_spun_node} with OCS storage label")
# DC app pods on the failed node will get automatically created on other
# running node. Waiting for all dc app pod to reach running state
pod.wait_for_dc_app_pods_to_reach_running_state(dc_pod_obj,
timeout=1200)
log.info("All the dc pods reached running state")
pod.wait_for_storage_pods()
# Check basic cluster functionality by creating resources
# (pools, storageclasses, PVCs, pods - both CephFS and RBD),
# run IO and delete the resources
self.sanity_helpers.create_resources(pvc_factory, pod_factory,
bucket_factory,
rgw_bucket_factory)
self.sanity_helpers.delete_resources()
# Perform cluster and Ceph health checks
self.sanity_helpers.health_check()
def test_automated_recovery_from_failed_nodes_IPI_proactive(
self, interface, pvc_factory, pod_factory, dc_pod_factory):
"""
Knip-678 Automated recovery from failed nodes
Proactive case - IPI
"""
# Get OSD running nodes
osd_running_nodes = get_osd_running_nodes()
log.info(f"OSDs are running on nodes {osd_running_nodes}")
# Label osd nodes with fedora app
label_worker_node(osd_running_nodes,
label_key='dc',
label_value='fedora')
# Create DC app pods
log.info("Creating DC based app pods")
interface = (constants.CEPHBLOCKPOOL
if interface == 'rbd' else constants.CEPHFILESYSTEM)
dc_pod_obj = []
for i in range(2):
dc_pod = dc_pod_factory(interface=interface,
node_selector={'dc': 'fedora'})
pod.run_io_in_bg(dc_pod, fedora_dc=True)
dc_pod_obj.append(dc_pod)
# Get app pods running nodes
dc_pod_node_name = get_app_pod_running_nodes(dc_pod_obj)
log.info(f"DC app pod running nodes are {dc_pod_node_name}")
# Get both osd and app pod running node
common_nodes = get_both_osd_and_app_pod_running_node(
osd_running_nodes, dc_pod_node_name)
msg = "Common OSD and app running node(s) NOT found"
assert (len(common_nodes) > 0), msg
log.info(f"Common OSD and app pod running nodes are {common_nodes}")
# Get the machine name using the node name
machine_name = machine.get_machine_from_node_name(common_nodes[0])
log.info(f"{common_nodes[0]} associated machine is {machine_name}")
# Get the machineset name using machine name
machineset_name = machine.get_machineset_from_machine_name(
machine_name)
log.info(
f"{common_nodes[0]} associated machineset is {machineset_name}")
# Add a new node and label it
add_new_node_and_label_it(machineset_name)
# Delete the machine
machine.delete_machine(machine_name)
log.info(f"Successfully deleted machine {machine_name}")
# DC app pods on the failed node will get automatically created on
# other running node. Waiting for all dc app pod to reach running
# state
pod.wait_for_dc_app_pods_to_reach_running_state(dc_pod_obj)
log.info("All the dc pods reached running state")
# Check all OCS pods status, they should be in running state
all_pod_obj = pod.get_all_pods(
namespace=defaults.ROOK_CLUSTER_NAMESPACE)
for pod_obj in all_pod_obj:
if ('-1-deploy' or 'ocs-deviceset') not in pod_obj.name:
try:
helpers.wait_for_resource_state(
resource=pod_obj,
state=constants.STATUS_RUNNING,
timeout=200)
except ResourceWrongStatusException:
# 'rook-ceph-crashcollector' on the failed node stucks at
# pending state. BZ 1810014 tracks it.
# Ignoring 'rook-ceph-crashcollector' pod health check as
# WA and deleting its deployment so that the pod
# disappears. Will revert this WA once the BZ is fixed
if 'rook-ceph-crashcollector' in pod_obj.name:
ocp_obj = ocp.OCP(
namespace=defaults.ROOK_CLUSTER_NAMESPACE)
pod_name = pod_obj.name
deployment_name = '-'.join(pod_name.split("-")[:-2])
command = f"delete deployment {deployment_name}"
ocp_obj.exec_oc_cmd(command=command)
log.info(f"Deleted deployment for pod {pod_obj.name}")
else:
raise
# Check basic cluster functionality by creating resources
# (pools, storageclasses, PVCs, pods - both CephFS and RBD),
# run IO and delete the resources
self.sanity_helpers.create_resources(pvc_factory, pod_factory)
self.sanity_helpers.delete_resources()
# Perform cluster and Ceph health checks
self.sanity_helpers.health_check()
def test_node_replacement_reactive_aws_ipi(
self, nodes, pvc_factory, pod_factory, dc_pod_factory,
failure, interface
):
"""
Knip-894 Node replacement - AWS-IPI-Reactive
"""
# Get worker nodes
initial_nodes = get_worker_nodes()
# Get OSD running nodes
osd_running_nodes = get_osd_running_nodes()
log.info(f"OSDs are running on nodes {osd_running_nodes}")
# Label osd nodes with fedora app
label_worker_node(osd_running_nodes, label_key='dc', label_value='fedora')
# Create DC app pods
log.info("Creating DC based app pods")
if interface == 'rbd':
interface = constants.CEPHBLOCKPOOL
elif interface == 'cephfs':
interface = constants.CEPHFILESYSTEM
dc_pod_obj = []
for i in range(2):
dc_pod = dc_pod_factory(
interface=interface, node_selector={'dc': 'fedora'})
pod.run_io_in_bg(dc_pod, fedora_dc=True)
dc_pod_obj.append(dc_pod)
# Get app pods running nodes
dc_pod_node_name = get_app_pod_running_nodes(dc_pod_obj)
log.info(f"DC app pod running nodes are {dc_pod_node_name}")
# Get both osd and app pod running node
common_nodes = get_both_osd_and_app_pod_running_node(
osd_running_nodes, dc_pod_node_name
)
log.info(f"Both OSD and app pod is running on nodes {common_nodes}")
# Get the machine name using the node name
machine_name = machine.get_machine_from_node_name(common_nodes[0])
log.info(f"{common_nodes[0]} associated machine is {machine_name}")
# Get the machineset name using machine name
machineset_name = machine.get_machineset_from_machine_name(
machine_name
)
log.info(
f"{common_nodes[0]} associated machineset is {machineset_name}"
)
# Get the failure node obj
failure_node_obj = get_node_objs(node_names=[common_nodes[0]])
# Induce failure on the selected failure node
log.info(f"Inducing failure on node {failure_node_obj[0].name}")
if failure == "power off":
# Power off AWS worker node instance
nodes.stop_nodes(failure_node_obj, wait=True)
log.info(f"Successfully powered off node: {failure_node_obj[0].name}")
elif failure == "network failure":
# Induce Network failure
node_network_failure([failure_node_obj[0].name])
# Add annotation to the failed node
annotation = "machine.openshift.io/exclude-node-draining=''"
machine.add_annotation_to_machine(
annotation=annotation, machine_name=machine_name
)
# Delete the machine
machine.delete_machine(machine_name)
log.info(f"Successfully deleted machine {machine_name}")
# Wait for the new machine to spin
log.info("Waiting for the new node to be in ready state")
machine.wait_for_new_node_to_be_ready(machineset_name)
# Get the node name of new spun node
nodes_after_new_spun_node = get_worker_nodes()
new_spun_node = list(
set(nodes_after_new_spun_node) - set(initial_nodes)
)
log.info(f"New spun node is {new_spun_node}")
# Label it
node_obj = ocp.OCP(kind='node')
node_obj.add_label(
resource_name=new_spun_node[0],
label=constants.OPERATOR_NODE_LABEL
)
log.info(
f"Successfully labeled {new_spun_node} with OCS storage label"
)
# DC app pods on the failed node will get automatically created on other
# running node. Waiting for all dc app pod to reach running state
pod.wait_for_dc_app_pods_to_reach_running_state(
dc_pod_obj, timeout=1200
)
log.info("All the dc pods reached running state")
# Check all OCS pods status, they should be in running state
all_pod_obj = pod.get_all_pods(
namespace=defaults.ROOK_CLUSTER_NAMESPACE
)
for pod_obj in all_pod_obj:
if '-1-deploy' and 'ocs-deviceset' not in pod_obj.name:
try:
helpers.wait_for_resource_state(
resource=pod_obj, state=constants.STATUS_RUNNING,
timeout=1800
)
except ResourceWrongStatusException:
# 'rook-ceph-crashcollector' on the failed node stucks at
# pending state. BZ 1810014 tracks it.
# Ignoring 'rook-ceph-crashcollector' pod health check as
# WA and deleting its deployment so that the pod
# disappears. Will revert this WA once the BZ is fixed
if 'rook-ceph-crashcollector' in pod_obj.name:
ocp_obj = ocp.OCP(
namespace=defaults.ROOK_CLUSTER_NAMESPACE
)
pod_name = pod_obj.name
deployment_name = '-'.join(pod_name.split("-")[:-2])
command = f"delete deployment {deployment_name}"
ocp_obj.exec_oc_cmd(command=command)
log.info(f"Deleted deployment for pod {pod_obj.name}")
# Check basic cluster functionality by creating resources
# (pools, storageclasses, PVCs, pods - both CephFS and RBD),
# run IO and delete the resources
self.sanity_helpers.create_resources(pvc_factory, pod_factory)
self.sanity_helpers.delete_resources()
# Perform cluster and Ceph health checks
self.sanity_helpers.health_check()