def login_all_targets(self,
udev_timeout=_UDEV_TIMEOUT,
iface_name="default"):
"""
Do an iSCSI login to all discovered targets
Args:
udev_timeout: (int) the default udev timeout is 120 seconds,
this can become a problem with high volumes of
"""
targets = [
target for target in DISCOVERED_TARGETS[self.system.name]
if target.iface == iface_name
]
log.debug("{}, iface: {}, targets: {}, DISCOVERED_TARGETS: {}".format(
self.system.name, iface_name, len(targets),
len(DISCOVERED_TARGETS[self.system.name])))
if len(targets) == 0:
msg = "Num of discovered targets with iface_name {} is 0".format(
iface_name)
log.warning(msg)
funcs_list = []
args_list = []
for target in DISCOVERED_TARGETS[self.system.name]:
funcs_list.append(self.login)
args_list.append((target.iqn, target.ip, 3260, target.iface))
pt = Parallel(funcs=funcs_list,
args_list=args_list,
max_workers=self.max_iscsi_worker)
pt.run_threads()
cmd = "sleep 1 ; udevadm trigger ; sleep 1 ;" \
" udevadm settle --timeout=%s" % udev_timeout
with _UDEV_SEMIS[self.system.name]:
self.system.run_cmd_check(cmd)
def wait_for_all_nodes_online(self):
"""
Waits until all nodes in the cluster are in running or recovered.
"""
nodes = self._clusterutil.list_nodes()
funcs_list = list()
args_list = list()
for node in nodes:
funcs_list.append(self.wait_for_node_online)
args_list.append([node])
parent = Parallel(funcs=funcs_list, args_list=args_list)
parent.run_threads()
logger.info("All nodes are online")
def rotate_logs(self):
"""
Force log rotation on all clients
"""
def _force_log_rotation_on_client(client):
logger.info("Forcing log rotation on client {}".format(
client.name))
client.force_log_rotation()
funcs = list()
args = list()
for client in self.clients:
funcs.append(_force_log_rotation_on_client)
args.append([client])
parent = Parallel(funcs=funcs, args_list=args, max_workers=len(funcs))
parent.run_threads()
def upload(self, logdir):
# TODO kill this if it takes too long! exec_command timeout?
self._logger.info("Beginning cluster log upload. This may take some time.")
fns = [self._process_node for _ in self._whitebox_list]
args_list = [[node, logdir] for node in self._whitebox_list]
Parallel(fns,
args_list=args_list,
max_workers=len(self._whitebox_list)).run_threads()
def upload(self, testcase):
if not self._client_list:
logger.debug("No client systems found, no nothing to upload")
return
fns = [self._process_client for _ in self._client_list]
args_list = [[client, testcase] for client in self._client_list]
Parallel(fns,
args_list=args_list,
max_workers=len(self._client_list)).run_threads()
def make_parallel(fn, args, results):
def run_fn(fn, arg, result_list):
if fn(arg) is False:
result_list.append(arg)
return
results = list()
funcs = [run_fn for arg in args]
args = [(fn, arg, results) for arg in args]
return Parallel(funcs, args_list=args)
def stop(self):
""" Stop all IO objects """
fn_list = []
for io in self._io_list:
fn_list.append(io.stop)
self._should_run = False
if self._thread is not None:
self._thread.join()
Parallel(fn_list).run_threads()
self.__io_has_been_stopped = True
def _log_time(self):
"""
As a means of keep time straight we want to log the current view of
time from everything that may have logs generated. This includes:
Executors, where the test is running.
Clients, where the IO is generated from.
"""
funcs = list()
args = list()
funcs.append(self._log_executor_time)
funcs.append(self._log_all_clients_time)
Parallel(funcs=funcs, max_workers=len(funcs)).run_threads()
def clients_can_reach_ips(opts):
"""
Threads each Client for all IPS
"""
ips = get_access_vips(opts)
if opts.all_ips:
ips.extend(get_all_ips(opts))
funcs = list()
args = list()
for client in qalib.client.list_from_equipment(opts.equipment,
required=False):
funcs.append(can_client_reach_ips)
args.append([client, ips])
parent = Parallel(funcs=funcs, args_list=args)
try:
parent.run_threads()
return True
except ValueError:
cluster = opts.equipment.get_cluster()
describe_cluster(cluster)
return False
def destroy_all_ifaces(self):
"""Destroys all ifaces that were created on the host."""
funcs = list()
args = list()
non_deletable_ifaces = ["default", "iser"]
for iface in self.list_ifaces():
if iface in non_deletable_ifaces:
continue
funcs.append(self.destroy_iface)
args.append([iface])
workers = min(16, len(funcs))
Parallel(funcs=funcs, args_list=args,
max_workers=workers).run_threads()
def logout_all_targets(self):
"""
Logout all targets which were logged into by this object
"""
# for iface scoped targets, lets do batch processing
# we will do this only for non default ifaces
if self.system.name in ISCSI_LOGGEDIN_TARGETS:
ifaces = []
for target in ISCSI_LOGGEDIN_TARGETS[self.system.name]:
if target.iface != "default":
ifaces.append(target.iface)
unique_ifaces = list(set(ifaces))
for iface in unique_ifaces:
# For unique interfaces, we can assume that only this client
# instance was used to do the iscsi login, so we can just log
# them all out with a single command:
cmd = ["iscsiadm", "-m node", "-I", iface, "-u"]
cmd_str = ' '.join(cmd)
with _ISCSIADM_SEMIS[self.system.name]:
ret, out = self.system.run_cmd(cmd_str)
# no error or exception raised, remove the entries from
# list
if ret == 0 and self.system.name in ISCSI_LOGGEDIN_TARGETS:
self.remove_target_from_iscsi_loggedin_targets(iface)
# Now do logouts individually for targets using the default
# interface. Run them in parallel
funcs_list = []
args_list = []
if (self.system.name in ISCSI_LOGGEDIN_TARGETS
and ISCSI_LOGGEDIN_TARGETS[self.system.name]):
for target in ISCSI_LOGGEDIN_TARGETS[self.system.name][:]:
funcs_list.append(self.logout)
args_list.append(
(target.iqn, target.ip, 3260, target.iface))
del ISCSI_LOGGEDIN_TARGETS[self.system.name]
pt = Parallel(funcs_list,
args_list,
max_workers=self.max_iscsi_worker)
pt.run_threads()
def _log_all_clients_time(self):
"""
Logs the client time in native time format, will need to be looked at
for windows or ESXi support.
"""
clients = qalib.client.list_from_equipment(self.equipment,
required=False)
funcs = list()
args = list()
for client in clients:
funcs.append(self._log_client_time)
args.append([client])
Parallel(funcs=funcs, args_list=args,
max_workers=len(funcs)).run_threads()
def configure_clients_from_client_si_map(self):
"""
Configure client with subset of si_list based on client_si_map
"""
# initialize initiator groups
self.add_client_initiators_to_storage_instances()
client_si_map = self.get_client_and_si_mapping()
_client_attach_funcs = []
_client_attach_args = []
success = False
try:
for client, si_list in client_si_map.viewitems():
# this is needed to handle fewer clients than SI's
if len(si_list) > 0:
_client_attach_funcs.append(
self._configure_clients_from_si_list)
_client_attach_args.append([client, si_list])
p = Parallel(_client_attach_funcs,
args_list=_client_attach_args,
max_workers=len(client_si_map.keys()))
p.run_threads()
success = True
finally:
if not success:
raise EnvironmentError(
"iscsi login failed from client {}: [{}]".format(
client.name, self._client_ifaces[client]))
else:
logger.debug(
"From client {}: [{}], iSCSI login succeed".format(
client.name, self._client_ifaces[client]))
return True
def ensure_cluster_ready(self):
"""
Run cluster health-checks required before beginning testing
"""
logger.info("Waiting for all nodes to be online")
self.cluster_util.health.wait_for_all_nodes_online()
logger.info("Running network diagnostics")
checks = ["interfaces", "ntp"]
for check in checks:
if not check_cluster_config_passes(
self.nodes[0], check, additional_flags=["parallel"]):
raise RuntimeError(
"check_cluster_config {} --parallel failed".format(check))
logger.info("Can clients reach VIPs?")
# TODO[jsp]: more of this can probably be
# moved into can_client_reach_ips
vips = list()
for vip in self.sdk.system.network.access_vip.get()["network_paths"]:
vips.append(vip["ip"])
if not vips:
raise RuntimeError("No VIPs detected on cluster")
funcs = list()
args = list()
for client in self.clients:
funcs.append(can_client_reach_ips)
args.append([client, vips])
parent = Parallel(funcs=funcs, args_list=args)
parent.run_threads()
logger.info("Can equipment be cleaned up?")
logger.info("...cleaning clients")
funcs = list()
args = list()
for client in self.clients:
funcs.append(lambda c: c.force_cleanup_all())
args.append([client])
continue
parent = Parallel(funcs=funcs,
args_list=args,
max_workers=len(self.clients))
parent.run_threads()
logger.info("...cleaning cluster")
self.cluster_util.force_clean()
logger.info("All cluster health checks completed successfully")
def start(self):
""" Start all IO objects """
if self.__io_has_been_started:
raise ValueError("start() cannot be called more than once")
fn_list = []
for io in self._io_list:
fn_list.append(io.start)
success = False
self._should_run = True
try:
Parallel(fn_list, max_workers=self.__max_workers).run_threads()
success = True
finally:
if not success:
self.stop()
self.__io_has_been_started = True
return self
def logout_all_sessions(self):
"""
Logs out of all sactive sessions on the host.
"""
funcs = list()
kwargs_list = list()
for session_info in self.list_session_info():
funcs.append(self.logout)
kwargs = {
"iqn": session_info["target_iqn"],
"ip": session_info["target_ip"],
"port": session_info["target_port"],
"iface_name": None
}
kwargs_list.append(kwargs)
workers = min(16, len(funcs))
Parallel(funcs=funcs, kwargs_list=kwargs_list,
max_workers=workers).run_threads()
def discover_all_targets(self,
targets,
iface_name="default",
iscsi_redirect=None):
"""
Performs iscsi discovery on a list of target addresses.
Parameter:
targets (list) - list of tuples with following values,
"ip, port, iqn"
iface_name (str) - interface name to limit discovery on
Raises:
EnvironmentError - If iqn is not discovered
"""
if iscsi_redirect is True:
# Save all the iqns from targets to tgt_iqns
# Pass tgt_iqns to discover()
tgt_iqns = []
for _, target in enumerate(targets):
(tgt_ip, tgt_port, tgt_iqn) = target
tgt_iqns.append(tgt_iqn)
self.discover(ip=tgt_ip,
iqn=tgt_iqns,
port=tgt_port,
iface_name=iface_name)
else:
funcs = list()
kwargs = list()
for _, target in enumerate(targets):
(tgt_ip, tgt_port, tgt_iqn) = target
funcs.append(self.discover)
kwargs.append({
"ip": tgt_ip,
"iqn": tgt_iqn,
"port": tgt_port,
"iface_name": iface_name
})
Parallel(funcs=funcs, kwargs_list=kwargs).run_threads()
def _thread_attach_all_targets(self, redirect_mode=None, redirect_ip=None):
"""Allocating a thread for each client to login."""
funcs = list()
kwargs = list()
for client in self._clients:
# update all targets with iface names before sending them out
# we need to get all targets
funcs.append(client.iscsi.attach_all_targets)
kwargs.append({
"targets": self._targets,
"iface_name": self._client_ifaces[client],
"redirect_mode": redirect_mode,
"redirect_ip": redirect_ip
})
Parallel(funcs=funcs,
kwargs_list=kwargs,
max_workers=len(self._clients)).run_threads()
for client in self._clients:
# TODO parallelize this block if needed
if not self._block_io:
client.filesystem.format_all_volumes(self.volumes)
client.filesystem.mount_all_volumes(self.volumes)
client.rescan_iscsi_bus()
def can_client_reach_ips(client, ips):
"""Will raise if client can not reach vip"""
failed_ips = list()
funcs = list()
args = list()
for ip in ips:
funcs.append(can_client_reach_ip)
args.append([client, ip, failed_ips])
Parallel(funcs=funcs, args_list=args).run_threads()
if failed_ips:
success_ips = (set(ips) -
set(failed_ips)).union(set(failed_ips) - set(ips))
msg = ("{}, is not able to ping {} ips, but is able to ping {} ips."
"failed IPs:\n".format(client.name, len(failed_ips),
len(success_ips)))
for ip in failed_ips:
msg += "{}\n".format(ip)
logger.error(msg)
raise ValueError(msg)
else:
msg = ("Client {} is able to ping {} ips.".format(
client.name, len(ips)))
logger.info(msg)
def main():
opts = get_args()
sys.excepthook = log_exceptions
setup_logging()
requests.packages.urllib3.disable_warnings(
requests.packages.urllib3.exceptions.InsecureRequestWarning)
logging.getLogger("requests").setLevel(logging.DEBUG)
logger.info("Reading cluster details...")
cluster = opts.equipment.get_cluster(required=True)
node_ips = cluster.get_server_ip_list()
nodes = map(
lambda n: wb.from_hostname(
n, username=cluster.admin_user, password=cluster.admin_password),
node_ips)
clients = qalib.client.list_from_equipment(opts.equipment, required=False)
# BEGIN PRE-CLUSTER SETUP
logger.info("Beginning pre-deploy checks")
logger.info("Are clients reachable?")
unpingable_clients = list()
parent = make_parallel(lambda c: c.is_pingable(), clients,
unpingable_clients)
parent.run_threads()
if unpingable_clients:
raise RuntimeError("Couldn't ping these client(s): {}".\
format(", ".join(unpingable_clients)))
logger.info("Are we able to login on client systems?")
def signin_client(client):
# this is basically client.run_cmd
shell = client._system_conn._get_shell()
try:
shell.exec_command("true")
return True
except paramiko.ssh_exception.AuthenticationException as _err:
return False
login_failed_clients = list()
parent = make_parallel(signin_client, clients, login_failed_clients)
parent.run_threads()
if login_failed_clients:
raise RuntimeError("Unable to login to client(s): {}".\
format(", ".join(login_failed_clients)))
logger.info("Installing required packages on clients")
funcs = list()
args = list()
for client in clients:
funcs.append(lambda c: c.setup())
args.append([client])
parent = Parallel(funcs=funcs, args_list=args)
parent.run_threads()
logger.info("Are storage nodes reachable?")
unpingable_nodes = list()
parent = make_parallel(lambda n: n.is_pingable(), nodes, unpingable_nodes)
parent.run_threads()
if unpingable_nodes:
raise RuntimeError("Unable to ping node(s): {}".format(
", ".join(unpingable_nodes)))
logger.info("Can we sign-in to all nodes as admin?")
failed_login_nodes = list()
def login_node(node):
ssh = SSH(node.ip,
username=cluster.admin_user,
password=cluster.admin_password)
try:
tp = ssh._open_transport()
tp.close()
return True
except paramiko.ssh_exception.AuthenticationException:
return False
parent = make_parallel(login_node, nodes, failed_login_nodes)
parent.run_threads()
if failed_login_nodes:
raise RuntimeError("Authentication failed: {}".format(
", ".join(failed_login_nodes)))
logger.info("Is REST interface up for all nodes?")
def is_rest_up(node, max_tries=3, wait_time=10):
tries = 0
path = "http://{}:8500".format(node.ip)
# TODO[jsp]: probably smarter to allow requests to handle max tries
# and backoff/wait time
while tries < max_tries:
tries += 1
try:
resp = requests.get(path)
if resp.status_code == 200:
return True
except requests.exceptions.ConnectTimeout as err:
logger.debug(str(err))
time.sleep(wait_time)
return False
rest_not_up = list()
parent = make_parallel(is_rest_up, nodes, rest_not_up)
parent.run_threads()
if rest_not_up:
raise RuntimeError("REST interface not up: {}".format(
", ".join(rest_not_up)))
logger.info("Verifying cluster network")
def check_ifaces(node):
return qalib.preflight.check_cluster_config_passes(node, "interfaces")
nodes_w_network_problems = list()
parent = make_parallel(check_ifaces, nodes, nodes_w_network_problems)
parent.run_threads()
logger.info("Pre-deploy work complete!")
logger.info("Deploying cluster")
deploy_cluster(opts.init, nodes[0])
if len(nodes) > 1:
funcs = list()
args = list()
requests_session = requests.Session()
retries = Retry(total=5, backoff_factor=0.1, status_forcelist=[111])
# TODO[jsp]: consider also using this adapter for init
adapter = HTTPAdapter(max_retries=retries)
requests_session.mount("http://", adapter)
for node in nodes[1:]:
funcs.append(join_cluster)
args.append([cluster.name, node, requests_session])
parent = Parallel(funcs=funcs, args_list=args)
parent.run_threads()
logger.info("Waiting for cluster to form")
sdk = get_cluster_api(cluster)
wait_for_cluster(sdk, node_ips)
logger.info("Cluster formed correctly.")
logger.info("Beginning post-deploy checks.")
preflight_helper = qalib.preflight.ClusterReady(opts.equipment)
preflight_helper.ensure_cluster_ready()
logger.info("Your cluster is ready.")
