def test_local_command2(self):
self.run_one_job(
# details = True,
job=SshJob(
node=LocalNode(),
commands=[
Run("for i in $(seq 3); do echo line $i; sleep 1; done"),
]))
def test_local_command(self):
# create random file in python rather than with /dev/random
# that is not working in virtualbox
random_full = "RANDOM-full"
random_head = "RANDOM"
self.random_file(random_full, size=19)
print("DONE")
self.run_one_job(
# details = True,
job=SshJob(
node=LocalNode(),
commands=[
Run(f"head -c {2**18} < {random_full} > {random_head}"),
Run(f"ls -l {random_head}"),
Run(f"shasum {random_head}"),
]))
def test_commands_verbose(self):
dummy_path = "tests/dummy-10"
dummy_file = Path(dummy_path).name
scheduler = Scheduler()
Sequence(SshJob(
node=self.gateway(),
verbose=True,
commands=[
Run("hostname"),
RunScript("tests/script-with-args.sh", "arg1", "arg2"),
RunString("for i in $(seq 3); do echo verbose$i; done"),
Push(localpaths=dummy_path, remotepath="."),
Pull(remotepaths=dummy_file, localpath=dummy_path + ".loop"),
]),
SshJob(node=LocalNode(),
critical=True,
commands=Run("diff {x} {x}.loop".format(x=dummy_path),
verbose=True)),
scheduler=scheduler)
ok = scheduler.run()
ok or scheduler.debrief()
self.assertTrue(ok)
nodes = (node1, node2)
########## create the scheduler instance upfront
scheduler = Scheduler()
check_lease = SshJob(
node=faraday,
critical=True,
command=Run("rhubarbe leases --check"),
scheduler=scheduler,
)
########## 1 step, generate a random data file of 1 M bytes
create_random_job = SshJob(
node=LocalNode(),
commands=[
Run("head", "-c", random_size, "<", "/dev/random", ">", "RANDOM"),
Run("ls", "-l", "RANDOM"),
Run("shasum", "RANDOM"),
],
required=check_lease,
scheduler=scheduler,
)
########## 2nd step : push this over to node1
push_job = SshJob(
node=node1,
commands=[
Push(localpaths=["RANDOM"], remotepath="."),
def main() -> bool:
parser = ArgumentParser(formatter_class=ArgumentDefaultsHelpFormatter)
parser.add_argument("-U", "--url", default=default_topurl,
dest='topurl',
help="url to reach nbhosting server")
parser.add_argument("-c", "--course-gitdir", default=default_course_gitdir,
help="""location of a git repo where to fetch notebooks;
needed in order to generate relevant URLs""")
parser.add_argument("-i", "--indices", default=[0], action=IntsRanges,
help="(cumulative) ranges of indices in the list of known notebooks"
" - run nbhtest with -l to see list")
parser.add_argument("-u", "--users", default=[1], action=IntsRanges,
help="(cumulative) ranges of students indexes; e.g. -u 101-400 -u 501-600")
parser.add_argument("-m", "--random", action='store_true',
help="if set, a random notebook index is used for each student")
parser.add_argument("-b", "--base", default='student',
help="basename for students name")
parser.add_argument("-p", "--period", default=20, type=float,
help="delay between 2 triggers of nbhtest")
parser.add_argument("-s", "--sleep", default=default_sleep_internal, type=float,
help="delay in seconds to sleep between actions inside nbhtest")
parser.add_argument("-w", "--window", default=default_window, type=int,
help="window depth for spawning the nbhtest instances")
parser.add_argument("-n", "--dry-run", action='store_true')
args = parser.parse_args()
course_gitdir = args.course_gitdir
course, notebooks = list_notebooks(course_gitdir)
# in random mode; what are the choices that we randomize on
if args.random:
if len(args.indices) > 1:
choices = args.indices
else:
choices = list(range(len(notebooks)))
local = LocalNode(
formatter=TerminalFormatter(
format="%H-%M-%S:@line@",
verbose=True
))
scheduler = Scheduler()
jobs = []
for user in args.users:
student_name = "{}-{:04d}".format(args.base, user)
if args.random:
indices = [ random.choice(choices) ]
else:
indices = args.indices
for index in indices:
command = "nbhtest.py -U {} -c {} -i {} -u {} -s {} &"\
.format(args.topurl, course_gitdir, index, student_name, args.sleep)
if args.dry_run:
print("dry-run:", command)
else:
# schule this command to run
job = Sequence(
SshJob(scheduler=scheduler,
node=local,
command = command,
),
Job(asyncio.sleep(args.period))
)
jobs.append(job)
if args.dry_run:
return True
overall = scheduler.orchestrate(
jobs_window = args.window
)
if not overall:
scheduler.debrief()
print("nbhtests DONE")
return overall
def run(*, gateway, slicename,
disaggregated_cn, operator_version, nodes, node_master, node_enb, quectel_nodes, phones, flexran,
drone, verbose, dry_run,
load_images, master_image, worker_image, quectel_image):
"""
Install K8S on R2lab
Arguments:
slicename: the Unix login name (slice name) to enter the gateway
quectel_nodes: list of indices of quectel UE nodes to use
phones: list of indices of phones to use
nodes: a list of node ids to run the scenario on; strings or ints
are OK;
node_master: the master node id, must be part of selected nodes
node_enb: the node id for the enb, which is connected to usrp/duplexer
disaggregated_cn: Boolean; True for the disaggregated CN scenario. False for all-in-one CN.
operator_version: str, either "none" or "v1" or "v2".
"""
if operator_version == "none":
only_kube5g = True
else:
only_kube5g = False
if node_master not in nodes:
print(f"master node {node_master} must be part of selected fit nodes {nodes}")
exit(1)
if node_enb not in nodes:
print(f"eNB worker node {node_enb} must be part of selected fit nodes {nodes}")
exit(1)
# Check if the browser can be automatically run to display the Drone app
if drone:
run_browser = True
if platform == "linux":
cmd_open = "xdg-open"
elif platform == "darwin":
cmd_open = "open"
else:
run_browser = False
if run_browser:
print(f"**** Will run the browser with command {cmd_open}")
else:
print(f"**** Will not be able to run the browser as platform is {platform}")
worker_ids = nodes[:]
worker_ids.remove(node_master)
quectel_ids = quectel_nodes[:]
quectel = len(quectel_ids) > 0
faraday = SshNode(hostname=default_gateway, username=slicename,
verbose=verbose,
formatter=TimeColonFormatter())
master = SshNode(gateway=faraday, hostname=fitname(node_master),
username="******",
verbose=verbose,
formatter=TimeColonFormatter())
node_index = {
id: SshNode(gateway=faraday, hostname=fitname(id),
username="******",formatter=TimeColonFormatter(),
verbose=verbose)
for id in nodes
}
nodes_quectel_index = {
id: SshNode(gateway=faraday, hostname=fitname(id),
username="******",formatter=TimeColonFormatter(),
verbose=verbose)
for id in quectel_nodes
}
worker_index = dict(node_index)
del worker_index[node_master]
fit_master = fitname(node_master)
fit_enb = fitname(node_enb)
# the global scheduler
scheduler = Scheduler(verbose=verbose)
##########
check_lease = SshJob(
scheduler=scheduler,
node = faraday,
critical = True,
verbose=verbose,
command = Run("rhubarbe leases --check"),
)
green_light = check_lease
if load_images:
green_light = [
SshJob(
scheduler=scheduler,
required=check_lease,
node=faraday,
critical=True,
verbose=verbose,
label = f"Load image {master_image} on master {fit_master}",
commands=[
Run(f"rhubarbe load {node_master} -i {master_image}"),
Run(f"rhubarbe wait {node_master}"),
]
),
SshJob(
scheduler=scheduler,
required=check_lease,
node=faraday,
critical=True,
verbose=verbose,
label = f"Load image {worker_image} on worker nodes",
commands=[
Run(f"rhubarbe usrpoff {node_enb}"), # if usrp is on, load could be problematic...
Run("rhubarbe", "load", *worker_ids, "-i", worker_image),
Run("rhubarbe", "wait", *worker_ids),
Run(f"rhubarbe usrpon {node_enb}"), # ensure a reset of the USRP on the enB node
],
),
SshJob(
scheduler=scheduler,
required=check_lease,
node=faraday,
critical=False,
verbose=verbose,
label="turning off unused nodes",
command=[
Run("rhubarbe bye --all "
+ "".join(f"~{x} " for x in nodes))
]
)
]
if quectel:
prepare_quectel = SshJob(
scheduler=scheduler,
required=green_light,
node=faraday,
critical=True,
verbose=verbose,
label = f"Load image {quectel_image} on quectel UE nodes",
commands=[
Run("rhubarbe", "usrpoff", *quectel_ids),
Run("rhubarbe", "load", *quectel_ids, "-i", quectel_image),
Run("rhubarbe", "wait", *quectel_ids),
Run("rhubarbe", "usrpon", *quectel_ids),
],
),
##########
if quectel:
# wait 30s for Quectel modules show up
wait_quectel_ready = PrintJob(
"Let Quectel modules show up",
scheduler=scheduler,
required=prepare_quectel,
sleep=30,
label="sleep 30s for the Quectel modules to show up"
)
# run the Quectel Connection Manager as a service on each Quectel UE node
quectelCM_service = Service(
command="quectel-CM -s oai.ipv4 -4",
service_id="QuectelCM",
verbose=verbose,
)
init_quectel_nodes = [
SshJob(
scheduler=scheduler,
required=wait_quectel_ready,
node=node,
critical=True,
verbose=verbose,
label=f"Init Quectel UE on fit node {id}",
commands = [
RunScript(find_local_embedded_script("nodes.sh"), "check-quectel-on", includes=INCLUDES),
quectelCM_service.start_command(),
],
) for id, node in nodes_quectel_index.items()
]
# wait 20s for Quectel Connection Manager to start up
wait_quectelCM_ready = PrintJob(
"Let QuectelCM start up",
scheduler=scheduler,
required=init_quectel_nodes,
sleep=20,
label="Sleep 20s for the Quectel Connection Manager(s) to start up"
)
detach_quectel_nodes = [
SshJob(
scheduler=scheduler,
required=wait_quectelCM_ready,
node=node,
critical=True,
verbose=verbose,
label=f"Detach Quectel UE on fit node {id}",
command = RunScript(find_local_embedded_script("nodes.sh"), "quectel-detach", includes=INCLUDES),
) for id, node in nodes_quectel_index.items()
]
##########
# Initialize k8s on the master node
init_master = SshJob(
scheduler=scheduler,
required=green_light,
node=master,
critical=True,
verbose=verbose,
label = f"Install and launch k8s on the master {node_master}",
commands = [
Run("swapoff -a"),
Run("hostnamectl set-hostname master-node"),
Run("kubeadm init --pod-network-cidr=10.244.0.0/16 > /tmp/join_msg.txt"),
Run("tail -2 /tmp/join_msg.txt > /tmp/join_msg"),
Run("mkdir -p $HOME/.kube"),
Run("cp -i /etc/kubernetes/admin.conf $HOME/.kube/config"),
Run("chown $(id -u):$(id -g) $HOME/.kube/config"),
Run("kubectl apply -f https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml"),
Run("kubectl get pods --all-namespaces"),
],
)
init_workers = [
SshJob(
scheduler=scheduler,
required=init_master,
node=node,
critical=True,
verbose=verbose,
label=f"Init k8s on fit node {id} and join the cluster",
commands = [
Run("swapoff -a"),
Run("increase-control-mtu"),
Run(f"scp -o 'StrictHostKeyChecking no' {fit_master}:/tmp/join_msg /tmp/join_msg"),
Run("chmod a+x /tmp/join_msg"),
Run("/tmp/join_msg"),
],
) for id, node in worker_index.items()
]
# wait 10s for K8S nodes setup
wait_k8nodes_ready = PrintJob(
"Let k8s set up",
scheduler=scheduler,
required=init_workers,
sleep=10,
label="sleep 10s for the k8s nodes to settle"
)
init_kube5g = SshJob(
scheduler=scheduler,
required = wait_k8nodes_ready,
node = master,
verbose=verbose,
label = f"Add oai:ran label to oai-ran pod on {node_enb} and start 5GOperator pod",
commands = [
Run("kubectl get nodes"),
# add label to the eNB node to help k8s scheduler selects the right fit node
Run(f"kubectl label nodes fit{node_enb} oai=ran"),
Run("kubectl get nodes -Loai"),
## retrieve the kube5g operator
#Run("git clone -b develop [email protected]:mosaic5g/kube5g.git"),
# install a few dependencies
Run("apt install -y python3-pip"),
Run("pip3 install --upgrade pip"),
Run("pip3 install ruamel.yaml==0.16.12 colorlog==4.6.2"),
Run("sed -i 's/oairan:v1-1.0-1/oairan:v1-1.0-3/g' /root/kube5g/common/config-manager/conf_global_default.yaml"),
# specify the R2lab specific configuration
Run("cd /root/kube5g/common/config-manager; ./conf-manager.py -s conf_short_r2lab.yaml"),
# apply the R2lab CRD
Run("cd /root/kube5g/openshift/kube5g-operator; ./k5goperator.sh -n"),
# start the kube5g operator pod
Run("cd /root/kube5g/openshift/kube5g-operator; ./k5goperator.sh container start"),
Run("kubectl get pods"),
],
)
# wait 30s for K8S 5G Operator setup
wait_k8_5GOp_ready = PrintJob(
"Let 5G Operator set up",
scheduler=scheduler,
required=init_kube5g,
sleep=30,
label="wait 30s for the 5G Operator pod to settle"
)
if only_kube5g:
finish = SshJob(
scheduler=scheduler,
required = wait_k8_5GOp_ready,
node = master,
verbose=verbose,
label = f"showing nodes and pods before leaving",
commands = [
Run("kubectl get nodes -Loai"),
Run("kubectl get pods"),
],
)
else:
if disaggregated_cn:
cn_type="disaggregated-cn"
# setup_time = 120
setup_time = 200
else:
cn_type="all-in-one"
# setup_time = 60
setup_time = 140
if flexran:
flexran_opt="flexran"
else:
flexran_opt=""
run_kube5g = SshJob(
scheduler=scheduler,
required = wait_k8_5GOp_ready,
node = master,
verbose=verbose,
label = f"deploy {operator_version} {cn_type} {flexran_opt} pods",
commands = [
Run("kubectl get nodes -Loai"),
Run(f"cd /root/kube5g/openshift/kube5g-operator; ./k5goperator.sh deploy {operator_version} {cn_type} {flexran_opt}"),
Run("kubectl get pods"),
],
)
# Coffee Break -- wait 1 or 2mn for K8S 5G pods setup
wait_k8_5Gpods_ready = PrintJob(
"Let all 5G pods set up",
scheduler=scheduler,
required=run_kube5g,
sleep=setup_time,
label=f"waiting {setup_time}s for all 5G pods to settle"
)
check_kube5g = SshJob(
scheduler=scheduler,
required = wait_k8_5Gpods_ready,
node = master,
verbose=verbose,
label = "Check which pods are deployed",
commands = [
Run("kubectl get nodes -Loai"),
Run("kubectl get pods"),
],
)
if drone:
# the place where runtime variables get stored
env = Variables()
#
# Define and run all the services to launch the Drone app locally on a firefox browser
#
drone_service = Service(
command=f"python /root/mosaic5g/store/sdk/frontend/drone/drone.py --port=8088 --tasks --address=192.168.3.{node_enb}",
service_id="drone_app",
verbose=verbose,
)
k8s_port9999_fwd_service = Service(
command=Deferred("kubectl port-forward {{flexran_pod}} 9999:9999 --address 0.0.0.0", env),
service_id="k8s-port9999-fwd",
verbose=verbose,
# somehow this is required for kubectl to run properly
environ={'KUBECONFIG': '/etc/kubernetes/admin.conf'}
)
# can't use a Service instance on the local box if it's not a Linux
# and we have macs...
local_port_fwd = (f"ssh -f -N -4"
f" -L9999:192.168.3.{node_master}:9999"
f" -L8088:192.168.3.{node_enb}:8088"
f" -o ExitOnForwardFailure=yes"
f" {slicename}@faraday.inria.fr")
browser_service = Service(
command=f"sleep 10; {cmd_open} http://127.0.0.1:8088/",
service_id="drone_browser",
verbose=verbose,
)
run_drone=SshJob(
scheduler=scheduler,
required=check_kube5g,
node=worker_index[node_enb],
verbose=verbose,
label=f"Run the drone app on worker node {node_enb} as a service",
commands=[
drone_service.start_command(),
],
)
get_flexran_podname=SshJob(
scheduler=scheduler,
required=check_kube5g,
node=master,
verbose=verbose,
label=f"Retrieve the name of the FlexRAN pod",
commands=[
# xxx here
Run("kubectl get --no-headers=true pods -l app=flexran -o custom-columns=:metadata.name",
capture=Capture('flexran_pod', env)),
],
)
run_k8s_port9999_fwd=SshJob(
scheduler=scheduler,
required=get_flexran_podname,
node=master,
verbose=verbose,
label=f"Run port forwarding on the master node as a service",
commands=[
k8s_port9999_fwd_service.start_command(),
],
)
# On the local machine, impossible to use Services as the latter uses systemd-run, only available on Linux
run_local_ports_fwd = SshJob(
scheduler=scheduler,
required = check_kube5g,
node = LocalNode(),
verbose=verbose,
label = f"Forward local ports 8088 and 9999",
command=Run(local_port_fwd + "&", ignore_outputs=True),
)
if run_browser:
run_local_browser = SshJob(
scheduler=scheduler,
required = (run_drone, run_k8s_port9999_fwd, run_local_ports_fwd),
node = LocalNode(),
verbose=verbose,
label = f"Run the browser on the local node in background",
command=browser_service.command+"&",
)
phones_requirements=run_local_browser
else:
phones_requirements=run_k8s_port9999_fwd
else:
phones_requirements=check_kube5g
########## Test phone(s) connectivity
sleeps_ran = (20, 25)
phone_msgs = [f"wait for {sleep}s for eNB to start up before waking up phone{id}"
for sleep, id in zip(sleeps_ran, phones)]
wait_commands = [f"echo {msg}; sleep {sleep}"
for msg, sleep in zip(phone_msgs, sleeps_ran)]
sleeps_phone = (15, 20)
phone2_msgs = [f"wait for {sleep}s for phone{id} before starting tests"
for sleep, id in zip(sleeps_phone, phones)]
wait2_commands = [f"echo {msg}; sleep {sleep}"
for msg, sleep in zip(phone2_msgs, sleeps_phone)]
job_start_phones = [
SshJob(
node=faraday,
commands=[
Run(wait_command),
RunScript(find_local_embedded_script("faraday.sh"), f"macphone{id}",
"r2lab-embedded/shell/macphone.sh", "phone-on",
includes=INCLUDES),
Run(wait2_command),
RunScript(find_local_embedded_script("faraday.sh"), f"macphone{id}",
"r2lab-embedded/shell/macphone.sh", "phone-check-cx",
includes=INCLUDES),
RunScript(find_local_embedded_script("faraday.sh"), f"macphone{id}",
"r2lab-embedded/shell/macphone.sh", "phone-start-app",
includes=INCLUDES),
],
label=f"turn off airplane mode on phone {id}",
required=phones_requirements,
scheduler=scheduler)
for id, wait_command, wait2_command in zip(phones, wait_commands, wait2_commands)
]
if quectel:
# wait 60s for Quectel connection(s) to set up
wait_before_attach_quectel = PrintJob(
"Wait again 30s before attaching Quectel device(s)",
scheduler=scheduler,
required=(job_start_phones,check_kube5g,detach_quectel_nodes),
sleep=30,
label="Sleep 30s before attaching Quectel device(s)"
)
job_attach_quectel = [
SshJob(
scheduler=scheduler,
required=wait_before_attach_quectel,
node=node,
critical=True,
verbose=verbose,
label=f"Attach Quectel UE on fit node {id}",
command = RunScript(find_local_embedded_script("nodes.sh"),
"quectel-attach", includes=INCLUDES),
) for id, node in nodes_quectel_index.items()
]
# wait 30s for Quectel connection(s) to set up
wait_quectel_cx_ready = PrintJob(
"Let the Quectel connection(s) set up",
scheduler=scheduler,
required=job_attach_quectel,
sleep=30,
label="Sleep 30s for the Quectel connection(s) to set up"
)
test_quectel_cx = [
SshJob(
scheduler=scheduler,
required=wait_quectel_cx_ready,
node=node,
critical=False,
verbose=verbose,
label=f"Check the Quectel cx on fit node {id}",
command = RunScript(find_local_embedded_script("nodes.sh"),
"check-quectel-cx", includes=INCLUDES),
) for id, node in nodes_quectel_index.items()
]
##########
# Update the .dot and .png file for illustration purposes
scheduler.check_cycles()
name = "deploy-kube5g"
print(10*'*', 'See main scheduler in',
scheduler.export_as_pngfile(name))
# orchestration scheduler jobs
if verbose:
scheduler.list()
if dry_run:
return True
if not scheduler.orchestrate():
print(f"RUN KO : {scheduler.why()}")
scheduler.debrief()
return False
print(f"RUN OK, you can log now on master node {fit_master} to manually change the scenario")
print(80*'*')
def one_run(*, protocol, interference,
run_name=default_run_name, slicename=default_slicename,
tx_power, phy_rate, antenna_mask, channel,
load_images=False,
node_ids=DEFAULT_NODE_IDS,
src_ids=DEFAULT_SRC_IDS, dest_ids=DEFAULT_DEST_IDS,
scrambler_id=DEFAULT_SCRAMBLER_ID,
tshark=False, map=False, warmup=False,
route_sampling=False, iperf=False,
verbose_ssh=False, verbose_jobs=False, dry_run=False,
run_number=None):
"""
Performs data acquisition on all nodes with the following settings
Arguments:
tx_power: in dBm, a string like 5, 10 or 14.
Corresponds to the transmission power.
phy_rate: a string among 1, 54. Correspond to the wifi rate.
antenna_mask: a string among 1, 3, 7.
channel: a string like e.g. 1 or 40. Correspond to the channel.
protocol: a string among batman , olsr. Correspond to the protocol
interference : in amplitude percentage, a string like 15 or 20.
Correspond to the power of the noise generated in the spectrum.
Can be either None or "None" to mean no interference.
run_name: the name for a subdirectory where all data will be kept
successive runs should use the same name for further visualization
slicename: the Unix login name (slice name) to enter the gateway
load_images: a boolean specifying whether nodes should be re-imaged first
node_ids: a list of node ids to run the scenario against;
strings or ints are OK;
tshark: a boolean specifying wether we should format/parse the .pcap.
map: a boolean specifying wether we should fetch/parse
the route tables of the nodes.
warmup: a boolean specifying whether we should run a ping before
the experiment to be certain of the stabilisation on the network.
src_ids: a list of nodes from which we will launch the ping from.
strings or ints are OK.
ping_messages : the number of ping packets that will be generated
"""
# set default for the nodes parameter
node_ids = ([int(id) for id in node_ids]
if node_ids is not None else DEFAULT_NODE_IDS)
src_ids = ([int(id) for id in src_ids]
if src_ids is not None else DEFAULT_SRC_IDS)
dest_ids = ([int(id) for id in dest_ids]
if dest_ids is not None else DEFAULT_NODE_IDS)
# all nodes - i.e. including sources and destinations -
# need to run the protocol
node_ids = list(set(node_ids).union(set(src_ids).union(set(dest_ids))))
if interference == "None":
interference = None
# open result dir no matter what
run_root = naming_scheme(
run_name=run_name, protocol=protocol,
interference=interference, autocreate=True)
# fix me trace = run_root / f"trace-{%m-%d-%H-%M}"
ref_time = apssh_time()
trace = run_root / f"trace-{ref_time}"
try:
with trace.open('w') as feed:
def log_line(line):
time_line(line, file=feed)
load_msg = f"{'WITH' if load_images else 'NO'} image loading"
interference_msg = (f"interference={interference} "
f"from scrambler={scrambler_id}")
nodes = " ".join(str(n) for n in node_ids)
srcs = " ".join(str(n) for n in src_ids)
dests = " ".join(str(n) for n in dest_ids)
ping_labels = [
f"PING {s} ➡︎ {d}"
for s in src_ids
# and on the destination
for d in dest_ids
if d != s
]
log_line(f"output in {run_root}")
log_line(f"trace in {trace}")
log_line(f"protocol={protocol}")
log_line(f"{load_msg}")
log_line(f"{interference_msg}")
log_line("----")
log_line(f"Selected nodes : {nodes}")
log_line(f"Sources : {srcs}")
log_line(f"Destinations : {dests}")
for label in ping_labels:
log_line(f"{label}")
log_line("----")
for feature in ('warmup', 'tshark', 'map',
'route_sampling', 'iperf'):
log_line(f"Feature {feature}: {locals()[feature]}")
except Exception as exc:
print(f"Cannot write into {trace} - aborting this run")
print(f"Found exception {type(exc)} - {exc}")
return False
#
# dry-run mode
# just display a one-liner with parameters
#
prelude = "" if not dry_run else "dry_run:"
with trace.open() as feed:
print(f"**************** {ref_time} one_run #{run_number}:")
for line in feed:
print(prelude, line, sep='', end='')
if dry_run:
return True
# the nodes involved
faraday = SshNode(hostname=default_gateway, username=slicename,
formatter=TimeColonFormatter(), verbose=verbose_ssh)
# this is a python dictionary that allows to retrieve a node object
# from an id
node_index = {
id: SshNode(gateway=faraday, hostname=fitname(id), username="******",
formatter=TimeColonFormatter(), verbose=verbose_ssh)
for id in node_ids
}
# extracts for sources and destinations
src_index = {id:node for (id, node) in node_index.items()
if id in src_ids}
dest_index = {id:node for (id, node) in node_index.items()
if id in dest_ids}
if interference:
node_scrambler = SshNode(
gateway=faraday, hostname=fitname(scrambler_id), username="******",
formatter=TimeColonFormatter(), verbose=verbose_ssh)
# the global scheduler
scheduler = Scheduler(verbose=verbose_jobs)
##########
check_lease = SshJob(
scheduler=scheduler,
node=faraday,
verbose=verbose_jobs,
label="rhubarbe check lease",
command=Run("rhubarbe leases --check", label="rlease"),
)
# load images if requested
green_light = check_lease
# at some point we did not load the scrambler if interference was None
# and that was a way to run faster loads with no interference
# but now we always load the scrambler node with gnuradio
# this is because when we do runs.py -i None 15 30 ...
# then the first call to one_run is with interference being None
# but it is still important to load the scrambler
if load_images:
# copy node_ids
load_ids = node_ids[:]
load_ids.append(scrambler_id)
# the nodes that we **do not** use should be turned off
# so if we have selected e.g. nodes 10 12 and 15, we will do
# rhubarbe off -a ~10 ~12 ~15, meaning all nodes except 10, 12 and 15
negated_node_ids = [f"~{id}" for id in load_ids]
# we can do these three things in parallel
ready_jobs = [
SshJob(node=faraday, required=green_light,
scheduler=scheduler, verbose=verbose_jobs,
command=Run("rhubarbe", "off", "-a", *negated_node_ids,
label="turn off unused nodes")),
SshJob(node=faraday, required=green_light,
scheduler=scheduler, verbose=verbose_jobs,
label="load batman image",
command=Run("rhubarbe", "load", "-i",
"batman-olsr",
*node_ids,
label=f"load ubuntu on {node_ids}")),
SshJob(
node=faraday, required=green_light,
scheduler=scheduler, verbose=verbose_jobs,
label="load gnuradio image",
command=Run("rhubarbe", "load", "-i",
"batman-olsr-gnuradio",
scrambler_id,
label=f"load gnuradio on {scrambler_id}")),
]
# replace green_light in this case
green_light = SshJob(
node=faraday, required=ready_jobs,
scheduler=scheduler, verbose=verbose_jobs,
label="wait for nodes to come up",
command=Run("rhubarbe", "wait", *load_ids))
##########
# setting up the wireless interface on all nodes
#
# provide node-utilities with the ranges/units it expects
frequency = channel_frequency[int(channel)]
# tx_power_in_mBm not in dBm
tx_power_driver = tx_power * 100
#just in case somme services failed in the previous experiment
reset_failed_services_job = [
SshJob(
node=node,
verbose=verbose_jobs,
label="reset failed services",
command=Run("systemctl reset-failed",
label="reset-failed services"))
for id, node in node_index.items()
]
reset_failed_services = Scheduler(
*reset_failed_services_job,
scheduler=scheduler,
required=green_light,
verbose=verbose_jobs,
label="Reset failed services")
init_wireless_sshjobs = [
SshJob(
node=node,
verbose=verbose_jobs,
label=f"init {id}",
command=RunScript(
"node-utilities.sh",
f"init-ad-hoc-network-{WIRELESS_DRIVER}",
WIRELESS_DRIVER, "foobar", frequency, phy_rate,
antenna_mask, tx_power_driver,
label="init add-hoc network"),
)
for id, node in node_index.items()]
init_wireless_jobs = Scheduler(
*init_wireless_sshjobs,
scheduler=scheduler,
required=green_light,
verbose=verbose_jobs,
label="Initialisation of wireless chips")
if interference:
# Run uhd_siggen with the chosen power
init_scrambler_job = SshJob(
scheduler=scheduler,
required=green_light,
forever=True,
node=node_scrambler,
verbose=verbose_jobs,
#TODO : If exit-signal patch is done add exit-signal=["TERM"]
# to this run object and call uhd_siggen directly
commands=[RunScript("node-utilities.sh",
"init-scrambler",
label="init scrambler"),
Run(f"systemd-run --unit=uhd_siggen -t ",
f"uhd_siggen -a usrp -f {frequency}M",
f"--sine --amplitude 0.{interference}",
label="systemctl start uhd_siggen")
]
)
green_light = [init_wireless_jobs, reset_failed_services]
# then install and run batman on fit nodes
run_protocol_job = [
SshJob(
# scheduler=scheduler,
node=node,
label=f"init and run {protocol} on fit node {id}",
verbose=verbose_jobs,
# CAREFUL : These ones use sytemd-run
# with the ----service-type=forking option!
command=RunScript("node-utilities.sh",
f"run-{protocol}",
label=f"run {protocol}"),
)
for id, node in node_index.items()]
run_protocol = Scheduler(
*run_protocol_job,
scheduler=scheduler,
required=green_light,
verbose=verbose_jobs,
label="init and run routing protocols")
green_light = run_protocol
# after that, run tcpdump on fit nodes, this job never ends...
if tshark:
run_tcpdump_job = [
SshJob(
# scheduler=scheduler_monitoring,
node=node,
forever=True,
label=f"run tcpdump on fit node {id}",
verbose=verbose_jobs,
command=[
Run("systemd-run -t --unit=tcpdump",
f"tcpdump -U -i moni-{WIRELESS_DRIVER}",
f"-y ieee802_11_radio -w /tmp/fit{id}.pcap",
label=f"tcpdump {id}")
]
)
for id, node in node_index.items()
]
run_tcpdump = Scheduler(
*run_tcpdump_job,
scheduler=scheduler,
required=green_light,
forever=True,
verbose=verbose_jobs,
label="Monitoring - tcpdumps")
# let the wireless network settle
settle_scheduler = Scheduler(
scheduler=scheduler,
required=green_light,
)
if warmup:
# warmup pings don't need to be sequential, so let's
# do all the nodes at the same time
# on a given node though, we'll ping the other ends sequentially
# see the graph for more
warmup_jobs = [
SshJob(
node=node_s,
verbose=verbose_jobs,
commands=[
RunScript("node-utilities.sh",
"my-ping", f"10.0.0.{d}",
warmup_ping_timeout,
warmup_ping_interval,
warmup_ping_size,
warmup_ping_messages,
f"warmup {s} ➡︎ {d}",
label=f"warmup {s} ➡︎ {d}")
for d in dest_index.keys()
if s != d
]
)
# for each selected experiment nodes
for s, node_s in src_index.items()
]
warmup_scheduler = Scheduler(
*warmup_jobs,
scheduler=settle_scheduler,
verbose=verbose_jobs,
label="Warmup pings")
settle_wireless_job2 = PrintJob(
"Let the wireless network settle after warmup",
sleep=settle_delay_shorter,
scheduler=settle_scheduler,
required=warmup_scheduler,
label=f"settling-warmup for {settle_delay_shorter} sec")
# this is a little cheating; could have gone before the bloc above
# but produces a nicer graphical output
# we might want to help asynciojobs if it offered a means
# to specify entry and exit jobs in a scheduler
settle_wireless_job = PrintJob(
"Let the wireless network settle",
sleep=settle_delay_long,
scheduler=settle_scheduler,
label=f"settling for {settle_delay_long} sec")
green_light = settle_scheduler
if iperf:
iperf_service_jobs = [
SshJob(
node=node_d,
verbose=verbose_jobs,
forever=True,
commands=[
Run("systemd-run -t --unit=iperf",
"iperf -s -p 1234 -u",
label=f"iperf serv on {d}"),
],
)
for d, node_d in dest_index.items()
]
iperf_serv_sched = Scheduler(
*iperf_service_jobs,
verbose=verbose_jobs,
label="Iperf Servers",
# for a nicer graphical output
# otherwise the exit arrow
# from scheduler 'iperf mode'
# to job 'settling for 60s'
# gets to start from this box
forever=True,
)
iperf_cli = [
SshJob(
node=node_s,
verbose=verbose_jobs,
commands=[
Run("sleep 7", label=""),
Run(f"iperf",
f"-c 10.0.0.{d} -p 1234",
f"-u -b {phy_rate}M -t 60",
f"-l 1024 > IPERF-{s:02d}-{d:02d}",
label=f"run iperf {s} ➡︎ {d}")
]
)
for s, node_s in src_index.items()
for d, node_d in dest_index.items()
if s != d
]
iperf_cli_sched = Scheduler(
Sequence(*iperf_cli),
verbose=verbose_jobs,
label="Iperf Clients")
iperf_stop = [
SshJob(node=node_d,
verbose=verbose_jobs,
label=f"Stop iperf on {d}",
command=Run("systemctl stop iperf"))
for d, node_d in dest_index.items()
]
iperf_stop_sched = Scheduler(
*iperf_stop,
required=iperf_cli_sched,
verbose=verbose_jobs,
label="Iperf server stop")
iperf_fetch = [
SshJob(node=node_s,
verbose=verbose_jobs,
command=Pull(
remotepaths=[f"IPERF-{s:02d}-{d:02d}"],
localpath=str(run_root),
label="fetch iperf {s} ➡︎ {d}")
)
for s, node_s in src_index.items()
for d, node_d in dest_index.items()
if s != d
]
iperf_fetch_sched = Scheduler(
*iperf_fetch,
required=iperf_stop_sched,
verbose=verbose_jobs,
label="Iperf fetch report")
iperf_jobs = [iperf_serv_sched, iperf_cli_sched,
iperf_stop_sched, iperf_fetch_sched]
iperf_sched = Scheduler(
*iperf_jobs,
scheduler=scheduler,
required=green_light,
verbose=verbose_jobs,
label="Iperf Module")
settle_wireless_job_iperf = PrintJob(
"Let the wireless network settle",
sleep=settle_delay_shorter,
scheduler=scheduler,
required=iperf_sched,
label=f"settling-iperf for {settle_delay_shorter} sec")
green_light = settle_wireless_job_iperf
# create all the tracepath jobs from the first node in the list
if map:
map_jobs = [
SshJob(
node=node,
label=f"Generating ROUTE file for proto {protocol} on node {id}",
verbose=verbose_jobs,
commands=[
RunScript(f"node-utilities.sh",
f"route-{protocol}",
f"> ROUTE-TABLE-{id:02d}",
label="get route table"),
Pull(remotepaths=[f"ROUTE-TABLE-{id:02d}"],
localpath=str(run_root),
label="")
],
)
for id, node in node_index.items()
]
map_scheduler = Scheduler(
*map_jobs,
scheduler=scheduler,
required=green_light,
verbose=verbose_jobs,
label="Snapshoting route files")
green_light = map_scheduler
if route_sampling:
route_sampling_jobs = [
SshJob(
node=node,
label=f"Route sampling service for proto {protocol} on node {id}",
verbose=False,
forever=True,
commands=[
Push(localpaths=["route-sample-service.sh"],
remotepath=".", label=""),
Run("chmod +x route-sample-service.sh", label=""),
Run("systemd-run -t --unit=route-sample",
"/root/route-sample-service.sh",
"route-sample",
f"ROUTE-TABLE-{id:02d}-SAMPLED",
protocol,
label="start route-sampling"),
],
)
for id, node in node_index.items()
]
route_sampling_scheduler = Scheduler(
*route_sampling_jobs,
scheduler=scheduler,
verbose=False,
forever=True,
label="Route Sampling services launch",
required=green_light)
##########
# create all the ping jobs, i.e. max*(max-1)/2
# this again is a python list comprehension
# see the 2 for instructions at the bottom
#
# notice that these SshJob instances are not yet added
# to the scheduler, we will add them later on
# depending on the sequential/parallel strategy
pings_job = [
SshJob(
node=node_s,
verbose=verbose_jobs,
commands=[
Run(f"echo actual ping {s} ➡︎ {d} using {protocol}",
label=f"ping {s} ➡︎ {d}"),
RunScript("node-utilities.sh", "my-ping",
f"10.0.0.{d}",
ping_timeout, ping_interval,
ping_size, ping_messages,
f"actual {s} ➡︎ {d}",
">", f"PING-{s:02d}-{d:02d}",
label=""),
Pull(remotepaths=[f"PING-{s:02d}-{d:02d}"],
localpath=str(run_root),
label=""),
],
)
# for each selected experiment nodes
for s, node_s in src_index.items()
for d, node_d in dest_index.items()
if s != d
]
pings = Scheduler(
scheduler=scheduler,
label="PINGS",
verbose=verbose_jobs,
required=green_light)
# retrieve all pcap files from fit nodes
stop_protocol_job = [
SshJob(
# scheduler=scheduler,
node=node,
# required=pings,
label=f"kill routing protocol on {id}",
verbose=verbose_jobs,
command=RunScript(f"node-utilities.sh",
f"kill-{protocol}",
label=f"kill-{protocol}"),
)
for id, node in node_index.items()
]
stop_protocol = Scheduler(
*stop_protocol_job,
scheduler=scheduler,
required=pings,
label="Stop routing protocols",
)
if tshark:
retrieve_tcpdump_job = [
SshJob(
# scheduler=scheduler,
node=nodei,
# required=pings,
label=f"retrieve pcap trace from fit{i:02d}",
verbose=verbose_jobs,
commands=[
Run("systemctl stop tcpdump",
label="stop tcpdump"),
#Run("systemctl reset-failed tcpdump"),
#RunScript("node-utilities.sh", "kill-tcpdump",
# label="kill-tcpdump"),
Run(
f"echo retrieving pcap trace and result-{i}.txt from fit{i:02d}",
label=""),
Pull(remotepaths=[f"/tmp/fit{i}.pcap"],
localpath=str(run_root), label=""),
],
)
for i, nodei in node_index.items()
]
retrieve_tcpdump = Scheduler(
*retrieve_tcpdump_job,
scheduler=scheduler,
required=pings,
label="Retrieve tcpdump",
)
if route_sampling:
retrieve_sampling_job = [
SshJob(
# scheduler=scheduler,
node=nodei,
# required=pings,
label=f"retrieve sampling trace from fit{i:02d}",
verbose=verbose_jobs,
commands=[
# RunScript("node-utilities.sh", "kill-route-sample", protocol,
# label = "kill route sample"),
#RunScript("route-sample-service.sh", "kill-route-sample",
# label="kill route sample"),
Run("systemctl stop route-sample",
label="stop route-sample"),
Run(
f"echo retrieving sampling trace from fit{i:02d}",
label=""),
Pull(remotepaths=[f"ROUTE-TABLE-{i:02d}-SAMPLED"],
localpath=str(run_root), label=""),
],
)
for i, nodei in node_index.items()
]
retrieve_sampling = Scheduler(
*retrieve_sampling_job,
scheduler=scheduler,
required=pings,
verbose=verbose_jobs,
label="Stop & retrieve route sampling",
)
if tshark:
parse_pcaps_job = [
SshJob(
# scheduler=scheduler,
node=LocalNode(),
# required=retrieve_tcpdump,
label=f"parse pcap trace {run_root}/fit{i}.pcap",
verbose=verbose_jobs,
#commands = [RunScript("parsepcap.sh", run_root, i)]
command=Run("tshark", "-2", "-r",
f"{run_root}/fit{i}.pcap",
"-R",
f"'(ip.dst==10.0.0.{i} && icmp) && radiotap.dbm_antsignal'",
"-Tfields",
"-e", "'ip.src'",
"-e" "'ip.dst'",
"-e", "'radiotap.dbm_antsignal'",
">", f"{run_root}/result-{i}.txt",
label=f"parsing pcap from {i}"),
)
for i in node_ids
]
parse_pcaps = Scheduler(
*parse_pcaps_job,
scheduler=scheduler,
required=retrieve_tcpdump,
label="Parse pcap",
)
if interference:
kill_uhd_siggen = SshJob(
scheduler=scheduler,
node=node_scrambler,
required=pings,
label=f"killing uhd_siggen on the scrambler node {scrambler_id}",
verbose=verbose_jobs,
commands=[Run("systemctl", "stop", "uhd_siggen"),
#Run("systemctl reset-failed tcpdump"),
],
)
kill_2_uhd_siggen = SshJob(
scheduler=scheduler,
node=faraday,
required=kill_uhd_siggen,
label=f"turning off usrp on the scrambler node {scrambler_id}",
verbose=verbose_jobs,
command=Run("rhubarbe", "usrpoff", scrambler_id),
)
pings.add(Sequence(*pings_job))
# for running sequentially we impose no limit on the scheduler
# that will be limitied anyways by the very structure
# of the required graph
# safety check
scheduler.export_as_pngfile(run_root / "experiment-graph")
if dry_run:
scheduler.list()
return True
# if not in dry-run mode, let's proceed to the actual experiment
ok = scheduler.run() # jobs_window=jobs_window)
# close all ssh connections
close_ssh_in_scheduler(scheduler)
# give details if it failed
if not ok:
scheduler.debrief()
scheduler.export_as_pngfile("debug")
if ok and map:
time_line("Creation of MAP files")
post_processor = ProcessRoutes(run_root, src_ids, node_ids)
post_processor.run()
if ok and route_sampling:
time_line("Creation of ROUTE SAMPLING files")
post_processor = ProcessRoutes(run_root, src_ids, node_ids)
post_processor.run_sampled()
# data acquisition is done, let's aggregate results
# i.e. compute averages
#if ok and tshark:
#post_processor = Aggregator(run_root, node_ids, antenna_mask)
#post_processor.run()
time_line("one_run done")
return ok
def run(
*,
# the pieces to use
slice,
hss,
epc,
enb,
phones,
e3372_ues,
oai_ues,
gnuradios,
e3372_ue_xterms,
oai_ue_xterms,
gnuradio_xterms,
# boolean flags
load_nodes,
skip_reset_usb,
oscillo,
# the images to load
image_gw,
image_enb,
image_oai_ue,
image_e3372_ue,
image_gnuradio,
# miscell
n_rb,
verbose,
dry_run):
"""
##########
# 3 methods to get nodes ready
# (*) load images
# (*) reset nodes that are known to have the right image
# (*) do nothing, proceed to experiment
expects e.g.
* slice : s.t like [email protected]
* hss : 04
* epc : 03
* enb : 23
* phones: list of indices of phones to use
* e3372_ues : list of nodes to use as a UE using e3372
* oai_ues : list of nodes to use as a UE using OAI
* gnuradios : list of nodes to load with a gnuradio image
* image_* : the name of the images to load on the various nodes
Plus
* load_nodes: whether to load images or not - in which case
image_gw, image_enb and image_*
are used to tell the image names
* skip_reset_usb : the USRP board will be reset as well unless this is set
"""
# what argparse knows as a slice actually is a gateway (user + host)
gwuser, gwhost = r2lab_parse_slice(slice)
gwnode = SshNode(hostname=gwhost,
username=gwuser,
formatter=TimeColonFormatter(verbose=verbose),
debug=verbose)
hostnames = hssname, epcname, enbname = [
r2lab_hostname(x) for x in (hss, epc, enb)
]
optional_ids = e3372_ues + oai_ues + gnuradios + \
e3372_ue_xterms + oai_ue_xterms + gnuradio_xterms
hssnode, epcnode, enbnode = [
SshNode(gateway=gwnode,
hostname=hostname,
username='******',
formatter=TimeColonFormatter(verbose=verbose),
debug=verbose) for hostname in hostnames
]
sched = Scheduler(verbose=verbose)
########## preparation
job_check_for_lease = SshJob(
node=gwnode,
command=["rhubarbe", "leases", "--check"],
label="check we have a current lease",
scheduler=sched,
)
# turn off all nodes
turn_off_command = ["rhubarbe", "off", "-a"]
# except our 3 nodes and the optional ones
turn_off_command += [
"~{}".format(x) for x in [hss, epc, enb] + optional_ids
]
# only do the turn-off thing if load_nodes
if load_nodes:
job_off_nodes = SshJob(
node=gwnode,
# switch off all nodes but the ones we use
command=turn_off_command,
label="turn off unused nodes",
required=job_check_for_lease,
scheduler=sched,
)
# actually run this in the gateway, not on the macphone
# the ssh keys are stored in the gateway and we do not yet have
# the tools to leverage such remote keys
job_stop_phones = [
SshJob(
node=gwnode,
command=RunScript(
# script
find_local_embedded_script("faraday.sh"),
# arguments
"macphone{}".format(id),
"r2lab-embedded/shell/macphone.sh",
"phone-off",
# options
includes=includes),
label="put phone{} in airplane mode".format(id),
required=job_check_for_lease,
scheduler=sched,
) for id in phones
]
########## prepare the image-loading phase
# this will be a dict of items imagename -> ids
to_load = defaultdict(list)
to_load[image_gw] += [hss, epc]
to_load[image_enb] += [enb]
if e3372_ues:
to_load[image_e3372_ue] += e3372_ues
if e3372_ue_xterms:
to_load[image_e3372_ue] += e3372_ue_xterms
if oai_ues:
to_load[image_oai_ue] += oai_ues
if oai_ue_xterms:
to_load[image_oai_ue] += oai_ue_xterms
if gnuradios:
to_load[image_gnuradio] += gnuradios
if gnuradio_xterms:
to_load[image_gnuradio] += gnuradio_xterms
prep_job_by_node = {}
for image, nodes in to_load.items():
commands = []
if load_nodes:
commands.append(Run("rhubarbe", "usrpoff", *nodes))
commands.append(Run("rhubarbe", "load", "-i", image, *nodes))
commands.append(Run("rhubarbe", "usrpon", *nodes))
# always do this
commands.append(Run("rhubarbe", "wait", "-t", 120, *nodes))
job = SshJob(
node=gwnode,
commands=commands,
label="Prepare node(s) {}".format(nodes),
required=job_check_for_lease,
scheduler=sched,
)
for node in nodes:
prep_job_by_node[node] = job
# start services
job_service_hss = SshJob(
node=hssnode,
command=RunScript(find_local_embedded_script("oai-hss.sh"),
"run-hss",
epc,
includes=includes),
label="start HSS service",
required=prep_job_by_node[hss],
scheduler=sched,
)
delay = 15
job_service_epc = SshJob(
node=epcnode,
commands=[
Run("echo giving HSS a headstart {delay}s to warm up; sleep {delay}"
.format(delay=delay)),
RunScript(find_local_embedded_script("oai-epc.sh"),
"run-epc",
hss,
includes=includes),
],
label="start EPC services",
required=prep_job_by_node[epc],
scheduler=sched,
)
########## enodeb
job_warm_enb = SshJob(
node=enbnode,
commands=[
RunScript(find_local_embedded_script("oai-enb.sh"),
"warm-enb",
epc,
n_rb,
not skip_reset_usb,
includes=includes),
],
label="Warm eNB",
required=prep_job_by_node[enb],
scheduler=sched,
)
enb_requirements = (job_warm_enb, job_service_hss, job_service_epc)
# wait for everything to be ready, and add an extra grace delay
grace = 30 if load_nodes else 10
grace_delay = SshJob(
node = LocalNode(formatter=TimeColonFormatter()),
command = "echo Allowing grace of {grace} seconds; sleep {grace}"\
.format(grace=grace),
required = enb_requirements,
scheduler = sched,
)
# start services
job_service_enb = SshJob(
node=enbnode,
# run-enb expects the id of the epc as a parameter
# n_rb means number of resource blocks for DL, set to either 25 or 50.
commands=[
RunScript(find_local_embedded_script("oai-enb.sh"),
"run-enb",
oscillo,
includes=includes,
x11=oscillo),
],
label="start softmodem on eNB",
required=grace_delay,
scheduler=sched,
)
########## run experiment per se
# Manage phone(s)
# this starts at the same time as the eNB, but some
# headstart is needed so that eNB actually is ready to serve
delay = 12
msg = "wait for {delay}s for enodeb to start up"\
.format(delay=delay)
wait_command = "echo {msg}; sleep {delay}".format(msg=msg, delay=delay)
job_start_phones = [
SshJob(
node=gwnode,
commands=[
Run(wait_command),
RunScript(find_local_embedded_script("faraday.sh"),
"macphone{}".format(id),
"r2lab-embedded/shell/macphone.sh",
"phone-on",
includes=includes),
RunScript(find_local_embedded_script("faraday.sh"),
"macphone{}".format(id),
"r2lab-embedded/shell/macphone.sh",
"phone-start-app",
includes=includes),
],
label="start Nexus phone and speedtest app",
required=grace_delay,
scheduler=sched,
) for id in phones
]
job_ping_phones_from_epc = [
SshJob(
node=epcnode,
commands=[
Run("sleep 10"),
Run("ping -c 100 -s 100 -i .05 172.16.0.{ip} &> /root/ping-phone"
.format(ip=id + 1)),
],
label="ping Nexus phone from EPC",
critical=False,
required=job_start_phones,
) for id in phones
]
########## xterm nodes
colors = ["wheat", "gray", "white", "darkolivegreen"]
xterms = e3372_ue_xterms + oai_ue_xterms + gnuradio_xterms
for xterm, color in zip(xterms, itertools.cycle(colors)):
xterm_node = SshNode(gateway=gwnode,
hostname=r2lab_hostname(xterm),
username='******',
formatter=TimeColonFormatter(verbose=verbose),
debug=verbose)
SshJob(
node=xterm_node,
command=Run("xterm -fn -*-fixed-medium-*-*-*-20-*-*-*-*-*-*-*"
" -bg {} -geometry 90x10".format(color),
x11=True),
label="xterm on node {}".format(xterm_node.hostname),
required=prep_job_by_node[xterm],
scheduler=sched,
# don't set forever; if we do, then these xterms get killed
# when all other tasks have completed
# forever = True,
)
# # remove dangling requirements - if any - should not be needed but won't hurt either
sched.sanitize()
print(20 * "*", "nodes usage summary")
if load_nodes:
for image, nodes in to_load.items():
for node in nodes:
print("node {node} : {image}".format(node=node, image=image))
else:
print("NODES ARE USED AS IS (no image loaded, no reset)")
print(10 * "*", "phones usage summary")
if phones:
for phone in phones:
print("Using phone{phone}".format(phone=phone))
else:
print("No phone involved")
sched.rain_check()
# Update the .dot and .png file for illustration purposes
if verbose or dry_run:
sched.list()
name = "scenario-load" if load_nodes else \
"scenario"
sched.export_as_dotfile("{name}.dot".format(name=name))
os.system("dot -Tpng {name}.dot -o {name}.png".format(name=name))
print("(Over)wrote {name}.png".format(name=name))
if dry_run:
return False
if verbose:
input('OK ? - press control C to abort ? ')
if not sched.orchestrate():
print("RUN KO : {}".format(sched.why()))
sched.debrief()
return False
else:
print("RUN OK")
return True
def one_run(tx_power,
phy_rate,
antenna_mask,
channel,
interference,
protocol,
*,
run_name=default_run_name,
slicename=default_slicename,
load_images=False,
node_ids=None,
verbose_ssh=False,
verbose_jobs=False,
dry_run=False,
tshark=False,
map=False,
warmup=False,
exp=default_exp,
dest=default_node_ids,
ping_number=default_ping_number,
route_sampling=False):
"""
Performs data acquisition on all nodes with the following settings
Arguments:
tx_power: in dBm, a string like 5, 10 or 14. Correspond to the transmission power.
phy_rate: a string among 1, 54. Correspond to the wifi rate.
antenna_mask: a string among 1, 3, 7.
channel: a string like e.g. 1 or 40. Correspond to the channel.
protocol: a string among batman , olsr. Correspond to the protocol
interference : in dBm, a string like 60 or 50. Correspond to the power of the noise generated in the root.
run_name: the name for a subdirectory where all data will be kept
successive runs should use the same name for further visualization
slicename: the Unix login name (slice name) to enter the gateway
load_images: a boolean specifying whether nodes should be re-imaged first
node_ids: a list of node ids to run the scenario against; strings or ints are OK;
defaults to the nodes [1, 4, 5, 12, 19, 22,27 ,31, 33, 37]
tshark: a boolean specifying wether we should format/parse the .pcap.
map: a boolean specifying wether we should fetch/parse the route tables of the nodes.
warmup: a boolean specifying wether we should run a ping before the experiment to be certain of the stabilisation on the network.
exp: a list of nodes from which we will launch the ping from. strings or ints are OK.
default to the node [1]
ping_number : The number of pings that will be generated
"""
# set default for the nodes parameter
node_ids = [int(id) for id in node_ids
] if node_ids is not None else default_node_ids
exp_ids = [int(id) for id in exp] if exp is not None else default_exp
dest_ids = [int(id)
for id in dest] if dest is not None else default_node_ids
#
# dry-run mode
# just display a one-liner with parameters
#
if dry_run:
print("************************************")
print("\n")
run_root = naming_scheme(protocol,
run_name,
tx_power,
phy_rate,
antenna_mask,
channel,
interference,
autocreate=False)
load_msg = "" if not load_images else " LOAD"
nodes = " ".join(str(n) for n in node_ids)
exps = " ".join(str(n) for n in exp)
pingst = [
"PING{}-->{}".format(e, j) for e in exp_ids
# and on the destination
for j in node_ids if e != j #and not
#(j in exp_ids and j < e)
]
print(
"dry-run:{protocol} {run_name}{load_msg} -"
" t{tx_power} r{phy_rate} a{antenna_mask} ch{channel} I{interference}-"
"nodes {nodes}"
" exp {exps}".format(**locals()))
print(
"\nNodes from which the experiment will be launched : \n{}\nList of pings generated:\n"
.format(exps))
print(pingst)
print("\n")
if warmup:
print("Will do warmup pings\n")
if tshark:
print(
"Will format data using tshark and will agregate the RSSI into one RSSI.txt file"
)
if map:
print(
"Will fetch the routing tables of the node (when stabilited) and will agregate the results\n"
)
if route_sampling:
print("Will launch route sampling services on nodes")
#print("Test creation of ROUTES files")
#post_processor= ProcessRoutes(run_root, exp_ids, node_ids)
#post_processor.run()
#print("\nList of tracepaths generated:\n{}".format(tracepathst))
# in dry-run mode we are done
###
# create the logs directory based on input parameters
run_root = naming_scheme(protocol,
run_name,
tx_power,
phy_rate,
antenna_mask,
channel,
interference,
autocreate=False)
if (run_root.is_dir()):
purgedir(run_root)
run_root = naming_scheme(protocol,
run_name,
tx_power,
phy_rate,
antenna_mask,
channel,
interference,
autocreate=True)
exp_info_file_name = run_root / "info.txt"
with exp_info_file_name.open("w") as info_file:
info_file.write("Selected nodes : \n")
for node in node_ids[:-1]:
info_file.write(f"{node} ")
info_file.write(f"{node_ids[-1]}")
info_file.write("\nSources : \n")
for src in exp_ids[:-1]:
info_file.write(f"{src} ")
info_file.write(f"{exp_ids[-1]}")
info_file.write("\nDestinations : \n")
for dest in dest_ids[:-1]:
info_file.write(f"{dest} ")
info_file.write(f"{dest_ids[-1]}" + "\n")
# the nodes involved
faraday = SshNode(hostname=default_gateway,
username=slicename,
formatter=TimeColonFormatter(),
verbose=verbose_ssh)
# this is a python dictionary that allows to retrieve a node object
# from an id
node_index = {
id: SshNode(gateway=faraday,
hostname=fitname(id),
username="******",
formatter=TimeColonFormatter(),
verbose=verbose_ssh)
for id in node_ids
}
if interference != "None":
node_scrambler = SshNode(gateway=faraday,
hostname=fitname(scrambler_id),
username="******",
formatter=TimeColonFormatter(),
verbose=verbose_ssh)
# the global scheduler
scheduler = Scheduler(verbose=verbose_jobs)
# if tshark:
#scheduler_monitoring = Scheduler(verbose=verbose_jobs)
#if interference != "None":
#scheduler_interferences = Scheduler(verbose=verbose_jobs)
##########
check_lease = SshJob(
scheduler=scheduler,
node=faraday,
verbose=verbose_jobs,
critical=True,
label="rhubarbe check lease",
command=Run("rhubarbe leases --check", label="rlease"),
#keep_connection = True
)
# load images if requested
green_light = check_lease
if load_images:
# the nodes that we **do not** use should be turned off
# so if we have selected e.g. nodes 10 12 and 15, we will do
# rhubarbe off -a ~10 ~12 ~15, meaning all nodes except 10, 12 and 15
negated_node_ids = ["~{}".format(id) for id in node_ids]
#Add the id of the scrambler in the list and load the gnuradio image
negated_node_ids.append("~{}".format(scrambler_id))
load_ids = [int(id) for id in node_ids
] if node_ids is not None else default_node_ids
load_ids.append(scrambler_id)
# replace green_light in this case
#We use a modified image of gnuradio where uhd_siggen handle the signal SIGTERM in order to finish properly
green_light = SshJob(
node=faraday,
required=check_lease,
#critical=True,
scheduler=scheduler,
verbose=verbose_jobs,
label="rhubarbe load/wait on nodes {}".format(load_ids),
commands=[
Run("rhubarbe",
"off",
"-a",
*negated_node_ids,
label="roff {}".format(negated_node_ids)),
Run("rhubarbe",
"load",
*node_ids,
label="rload {}".format(node_ids)),
Run("rhubarbe",
"load",
"-i",
"gnuradio_batman",
scrambler_id,
label="load gnuradio batman on {}".format(scrambler_id)),
Run("rhubarbe", "wait", *load_ids, label="rwait")
],
#keep_connection = True
)
##########
# setting up the wireless interface on all nodes
#
# this is a python feature known as a list comprehension
# we just create as many SshJob instances as we have
# (id, SshNode) couples in node_index
# and gather them all in init_wireless_jobs
# they all depend on green_light
#
# provide node-utilities with the ranges/units it expects
frequency = channel_frequency[int(channel)]
# tx_power_in_mBm not in dBm
tx_power_driver = tx_power * 100
init_wireless_sshjobs = [
SshJob(
#scheduler=scheduler,
#required=green_light,
node=node,
verbose=verbose_jobs,
label="init {}".format(id),
command=RunScript("node-utilities.sh",
"init-ad-hoc-network-{}".format(wireless_driver),
wireless_driver,
"foobar",
frequency,
phy_rate,
antenna_mask,
tx_power_driver,
label="init add-hoc network"),
#keep_connection = True
) for id, node in node_index.items()
]
init_wireless_jobs = Scheduler(
*init_wireless_sshjobs,
scheduler=scheduler,
required=green_light,
#critical = True,
verbose=verbose_jobs,
label="Initialisation of wireless chips")
green_light_prot = init_wireless_jobs
if interference != "None":
#Run uhd_siggen with the chosen power
frequency_str = frequency / 1000
frequency_str = str(frequency_str) + "G"
init_scrambler_job = [
SshJob(
forever=True,
node=node_scrambler,
verbose=verbose_jobs,
label="init scrambler on node {}".format(scrambler_id),
command=RunScript("node-utilities.sh",
"init-scrambler",
interference,
frequency_str,
label="init scambler"),
#keep_connection = True
)
]
init_scrambler = Scheduler(
*init_scrambler_job,
scheduler=scheduler,
required=green_light,
#forever = True,
#critical = True,
verbose=verbose_jobs,
label="Running interference")
# then install and run batman on fit nodes
run_protocol_job = [
SshJob(
#scheduler=scheduler,
node=node,
#required=green_light_prot,
label="init and run {} on fit node {}".format(protocol, i),
verbose=verbose_jobs,
command=RunScript("node-utilities.sh",
"run-{}".format(protocol),
label="run {}".format(protocol)),
#keep_connection = True
) for i, node in node_index.items()
]
run_protocol = Scheduler(
*run_protocol_job,
scheduler=scheduler,
required=green_light_prot,
#critical = True,
verbose=verbose_jobs,
label="init and run routing protocols")
# after that, run tcpdump on fit nodes, this job never ends...
if tshark:
run_tcpdump_job = [
SshJob(
#scheduler=scheduler_monitoring,
node=node,
forever=True,
label="run tcpdump on fit node".format(i),
verbose=verbose_jobs,
commands=[
RunScript("node-utilities.sh",
"run-tcpdump",
wireless_driver,
i,
label="run tcpdump")
],
#keep_connection = True
) for i, node in node_index.items()
]
run_tcpdump = Scheduler(
*run_tcpdump_job,
scheduler=scheduler,
required=run_protocol,
forever=True,
#critical = True,
verbose=verbose_jobs,
label="Monitoring (tcpdum) Jobs")
# let the wireless network settle
settle_wireless_job = PrintJob(
"Let the wireless network settle",
sleep=settle_delay,
scheduler=scheduler,
required=run_protocol,
label="settling for {} sec".format(settle_delay))
green_light_experiment = settle_wireless_job
if warmup:
warmup_pings_job = [
SshJob(
node=nodei,
#required=green_light_experiment,
label="warmup ping {} -> {}".format(i, j),
verbose=verbose_jobs,
commands=[
Run("echo {} '->' {}".format(i, j),
label="ping {} '->' {}".format(i, j)),
RunScript("node-utilities.sh",
"my-ping",
"10.0.0.{}".format(j),
ping_timeout,
ping_interval,
ping_size,
ping_number,
label="")
],
#keep_connection = True
)
#for each selected experiment nodes
for e in exp_ids
# looping on the source (to get the correct sshnodes)
for i, nodei in node_index.items()
# and on the destination
for j, nodej in node_index.items()
# and keep only sources that are in the selected experiment nodes and remove destination that are themselves
# and remove the couples that have already be done
# print("i {index} exp {expe}".format(index = i, expe= exp))
if (i == e) and e != j and not (j in exp_ids and j < e)
]
warmup_pings = Scheduler(
Sequence(*warmup_pings_job),
scheduler=scheduler,
required=green_light_experiment,
#critical = True,
verbose=verbose_jobs,
label="Warmup ping")
settle_wireless_job2 = PrintJob(
"Let the wireless network settle",
sleep=settle_delay / 2,
scheduler=scheduler,
required=warmup_pings,
label="settling-warmup for {} sec".format(settle_delay / 2))
green_light_experiment = settle_wireless_job2
##########
# create all the tracepath jobs from the first node in the list
#
if map:
routes_job = [
SshJob(
node=nodei,
#scheduler=scheduler,
#required=green_light_experiment,
label="Generating ROUTE file for prot {} on node {}".format(
protocol, i),
verbose=verbose_jobs,
commands=[
RunScript("node-utilities.sh",
"route-{}".format(protocol),
">",
"ROUTE-TABLE-{:02d}".format(i),
label="get route table"),
Pull(remotepaths="ROUTE-TABLE-{:02d}".format(i),
localpath=str(run_root),
label="")
],
#keep_connection = True
) for i, nodei in node_index.items()
]
routes = Scheduler(
*routes_job,
scheduler=scheduler,
required=green_light_experiment,
#critical = True,
verbose=verbose_jobs,
label="Snapshoting route files")
green_light_experiment = routes
if route_sampling:
routes_sampling_job2 = [
SshJob(
node=nodei,
label="Route sampling service for prot {} on node {}".format(
protocol, i),
verbose=False,
#forever = True,
commands=[
Push(localpaths=["route_sample_service.sh"],
remotepath=".",
label=""),
Run("source",
"route_sample_service.sh;",
"route-sample",
"ROUTE-TABLE-{:02d}-SAMPLED".format(i),
"{}".format(protocol),
label="run route sampling service"),
],
#keep_connection = True
) for i, nodei in node_index.items()
]
routes_sampling_job = [
SshJob(
node=nodei,
label="Route sampling service for prot {} on node {}".format(
protocol, i),
verbose=False,
forever=True,
#critical = True,
#required = green_light_experiment,
#scheduler = scheduler,
commands=[
RunScript("route_sample_service.sh",
"route-sample",
"ROUTE-TABLE-{:02d}-SAMPLED".format(i),
"{}".format(protocol),
label="run route sampling service"),
],
#keep_connection = True
) for i, nodei in node_index.items()
]
routes_sampling = Scheduler(
*routes_sampling_job,
scheduler=scheduler,
verbose=False,
forever=True,
#critical = True,
label="Route Sampling services launch",
required=green_light_experiment)
##########
# create all the ping jobs, i.e. max*(max-1)/2
# this again is a python list comprehension
# see the 2 for instructions at the bottom
#
# notice that these SshJob instances are not yet added
# to the scheduler, we will add them later on
# depending on the sequential/parallel strategy
pings_job = [
SshJob(
node=nodei,
#required=green_light_experiment,
label="ping {} -> {}".format(i, j),
verbose=verbose_jobs,
commands=[
Run("echo {} '->' {}".format(i, j),
label="ping {}'->' {}".format(i, j)),
RunScript("node-utilities.sh",
"my-ping",
"10.0.0.{}".format(j),
ping_timeout,
ping_interval,
ping_size,
ping_number,
">",
"PING-{:02d}-{:02d}".format(i, j),
label=""),
Pull(remotepaths="PING-{:02d}-{:02d}".format(i, j),
localpath=str(run_root),
label=""),
],
#keep_connection = True
)
#for each selected experiment nodes
for e in exp_ids
# looping on the source (to get the correct sshnodes)
for i, nodei in node_index.items()
# and on the destination
for j in dest_ids
# and keep only sources that are in the selected experiment nodes and remove destination that are themselves
# and remove the couples that have already be done
if (i == e) and e != j and not (j in exp_ids and j < e)
]
pings = Scheduler(
scheduler=scheduler,
label="PINGS",
#critical = True,
verbose=verbose_jobs,
required=green_light_experiment)
# retrieve all pcap files from fit nodes
stop_protocol_job = [
SshJob(
#scheduler=scheduler,
node=nodei,
#required=pings,
label="kill routing protocol on fit{:02d}".format(i),
verbose=verbose_jobs,
#critical = True,
commands=[
RunScript("node-utilities.sh",
"kill-{}".format(protocol),
label="kill-{}".format(protocol)),
],
#keep_connection = False
) for i, nodei in node_index.items()
]
stop_protocol = Scheduler(
*stop_protocol_job,
scheduler=scheduler,
required=pings,
#critical = True,
label="Stop routing protocols",
)
if tshark:
retrieve_tcpdump_job = [
SshJob(
#scheduler=scheduler,
node=nodei,
#required=pings,
label="retrieve pcap trace from fit{:02d}".format(i),
verbose=verbose_jobs,
#critical = True,
commands=[
# RunScript("node-utilities.sh", "kill-{}".format(protocol), label = "kill-{}".format(protocol)),
RunScript("node-utilities.sh",
"kill-tcpdump",
label="kill-tcpdump"),
#Run("sleep 1"),
Run("echo retrieving pcap trace and result-{i}.txt from fit{i:02d}"
.format(i=i),
label=""),
Pull(remotepaths=["/tmp/fit{}.pcap".format(i)],
localpath=str(run_root),
label=""),
],
#keep_connection = True
) for i, nodei in node_index.items()
]
retrieve_tcpdump = Scheduler(
*retrieve_tcpdump_job,
scheduler=scheduler,
required=pings,
#critical = True,
label="Retrieve tcpdump",
)
if route_sampling:
retrieve_sampling_job = [
SshJob(
#scheduler=scheduler,
node=nodei,
#required=pings,
label="retrieve sampling trace from fit{:02d}".format(i),
verbose=verbose_jobs,
#critical = True,
commands=[
#RunScript("node-utilities.sh", "kill-route-sample", protocol,
# label = "kill route sample"),
RunScript("route_sample_service.sh",
"kill-route-sample",
label="kill route sample"),
Run("echo retrieving sampling trace from fit{i:02d}".
format(i=i),
label=""),
Pull(remotepaths=["ROUTE-TABLE-{:02d}-SAMPLED".format(i)],
localpath=str(run_root),
label=""),
],
#keep_connection = True
) for i, nodei in node_index.items()
]
retrieve_sampling = Scheduler(
*retrieve_sampling_job,
scheduler=scheduler,
required=pings,
#critical=True,
verbose=verbose_jobs,
label="Retrieve & stopping route sampling",
)
if tshark:
parse_pcaps_job = [
SshJob(
#scheduler=scheduler,
node=LocalNode(),
#required=retrieve_tcpdump,
label="parse pcap trace {path}/fit{node}.pcap".format(
path=run_root, node=i),
verbose=verbose_jobs,
#commands = [RunScript("parsepcap.sh", run_root, i)]
commands=[
Run("tshark",
"-2",
"-r",
"{path}/fit{node}.pcap".format(path=run_root, node=i),
"-R",
"'(ip.dst==10.0.0.{node} && icmp) && radiotap.dbm_antsignal'"
.format(node=i),
"-Tfields",
"-e",
"'ip.src'", "-e"
"'ip.dst'",
"-e",
"'radiotap.dbm_antsignal'",
">",
"{path}/result-{node}.txt".format(path=run_root,
node=i),
label="parse pcap locally")
],
#keep_connection = True
) for i in node_ids
]
parse_pcaps = Scheduler(
*parse_pcaps_job,
scheduler=scheduler,
required=retrieve_tcpdump,
#critical=True,
label="Parse pcap",
)
#TODO: TURN OFF USRP
if interference != "None":
kill_uhd_siggen = SshJob(
scheduler=scheduler,
node=node_scrambler,
required=pings,
label="killing uhd_siggen on the scrambler node {}".format(
scrambler_id),
verbose=verbose_jobs,
#critical = True,
commands=[Run("pkill", "uhd_siggen")],
#keep_connection = True
)
kill_2_uhd_siggen = SshJob(
scheduler=scheduler,
node=faraday,
required=kill_uhd_siggen,
label="turning off usrp on the scrambler node {}".format(
scrambler_id),
verbose=verbose_jobs,
commands=[
Run("rhubarbe", "usrpoff", "fit{}".format(scrambler_id))
],
#keep_connection = True
)
#if map:
#scheduler.add(Sequence(*tracepaths, scheduler=scheduler))
#if warmup:
# scheduler.add(Sequence(*warmup_pings_job, scheduler=scheduler))
pings.add(Sequence(*pings_job))
# for running sequentially we impose no limit on the scheduler
# that will be limitied anyways by the very structure
# of the required graph
#jobs_window = None
if dry_run:
scheduler.export_as_pngfile(run_root / "experiment_graph")
return True
# if not in dry-run mode, let's proceed to the actual experiment
ok = scheduler.orchestrate() #jobs_window=jobs_window)
scheduler.shutdown()
dot_file = run_root / "experiment_graph"
if not dot_file.is_file():
scheduler.export_as_dotfile(dot_file)
#TODO : Is it necessary? if the user want to see it he can just do it?
#call(["dot", "-Tpng", dot_file, "-o", run_root / "experitment_graph.png"])
#ok=True
#ok = False
# give details if it failed
if not ok:
scheduler.debrief()
scheduler.export_as_dotfile("debug")
if ok and map:
print("Creation of ROUTES files")
post_processor = ProcessRoutes(run_root, exp_ids, node_ids)
post_processor.run()
if ok and route_sampling:
post_processor = ProcessRoutes(run_root, exp_ids, node_ids)
post_processor.run_sampled()
print("END of creation for ROUTES FILES")
# data acquisition is done, let's aggregate results
# i.e. compute averages
if ok and tshark:
post_processor = Aggregator(run_root, node_ids, antenna_mask)
post_processor.run()
return ok
def run(self, verbose, no_load, no_save):
"""
can skip the load or save phases
"""
print("Using node {} through gateway {}".format(
self.node, self.gateway))
print("In order to produce {} from {}".format(self.to_image,
self.from_image))
print("The following scripts will be run:")
for i, script in enumerate(self.scripts, 1):
print("{:03d}:{}".format(i, " ".join(script)))
items = []
if no_load: items.append("skip load")
if no_save: items.append("skip save")
if items:
print("WARNING: using fast-track mode {}".format(
' & '.join(items)))
self.locate_companion_shell()
if verbose:
print("Located companion in {}".format(self.companion))
if verbose:
print("Preparing tar of input shell scripts .. ", end="")
tarfile = self.prepare_tar(self.to_image)
if verbose:
print("Done in {}".format(tarfile))
keys = load_agent_keys()
if verbose:
print("We have found {} keys in the ssh agent".format(len(keys)))
#################### the 2 nodes we need to talk to
gateway_proxy = None
gwuser, gwname = self.user_host(self.gateway)
gateway_proxy = None if not gwuser else SshNode(
hostname=gwname,
username=gwuser,
keys=keys,
formatter=ColonFormatter(verbose=verbose),
)
# really not sure it makes sense to use username other than root
username, nodename = self.user_host(self.node)
node_proxy = SshNode(
gateway=gateway_proxy,
hostname=nodename,
username=username,
keys=keys,
formatter=ColonFormatter(verbose=verbose),
)
banner = 20 * '='
# now that node_proxy is initialized, we need to
# have a valid gateway_proxy for when we run all this from inside
# the gateway
if gateway_proxy is None:
print("WARNING: build-image is designed to be run on your laptop")
# best-effort, not even tested....
gateway_proxy = LocalNode()
#################### the little pieces
sequence = Sequence(
PrintJob("Checking for a valid lease"),
# bail out if we don't have a valid lease
SshJob(node = gateway_proxy,
command = "rhubarbe leases --check",
critical = True),
PrintJob("loading image {}".format(self.from_image)
if not no_load else "fast-track: skipping image load",
banner = banner,
# label = "welcome message",
),
SshJob(
node = gateway_proxy,
commands = [
Run("rhubarbe", "load", "-i", self.from_image, nodename) \
if not no_load else None,
Run("rhubarbe", "wait", "-v", "-t", "240", nodename),
],
# label = "load and wait image {}".format(self.from_image),
),
SshJob(
node = node_proxy,
commands = [
Run("rm", "-rf", "/etc/rhubarbe-history/{}".format(self.to_image)),
Run("mkdir", "-p", "/etc/rhubarbe-history"),
Push(localpaths = tarfile,
remotepath = "/etc/rhubarbe-history"),
RunScript(self.companion, nodename, self.from_image, self.to_image),
Pull(localpath = "{}/logs/".format(self.to_image),
remotepaths = "/etc/rhubarbe-history/{}/logs/".format(self.to_image),
recurse = True),
],
label = "set up and run scripts in /etc/rhubarbe-history/{}".format(self.to_image)),
)
# avoid creating an SshJob with void commands
if self.extra_logs:
sequence.append(
SshJob(
node=node_proxy,
label="collecting extra logs",
critical=False,
commands=[
Pull(localpath="{}/logs/".format(self.to_image),
remotepaths=extra_log,
recurse=True) for extra_log in self.extra_logs
],
))
# creating these as critical = True means the whole
# scenario will fail if these are not found
for binary in self.expected_binaries:
check_with = "ls" if os.path.isabs(binary) else ("type -p")
sequence.append(
Sequence(
PrintJob(
"Checking for expected binaries",
# label = "message about checking"
),
SshJob(
node=node_proxy,
command=[check_with, binary],
# label = "Checking for {}".format(binary)
)))
# xxx some flag
if no_save:
sequence.append(
PrintJob("fast-track: skipping image save", banner=banner))
else:
sequence.append(
Sequence(
PrintJob("saving image {} ...".format(self.to_image),
banner=banner),
# make sure we capture all the logs and all that
# mostly to test RunString
SshJob(
node=node_proxy,
command=RunString("sync ; sleep $1; sync; sleep $1",
1),
# label = 'sync',
),
SshJob(
node=gateway_proxy,
command=Run("rhubarbe", "save", "-o", self.to_image,
nodename),
# label = "save image {}".format(self.to_image),
),
SshJob(
node=gateway_proxy,
command="rhubarbe images -d",
# label = "list current images",
),
))
sched = Scheduler(sequence, verbose=verbose)
# sanitizing for the cases where some pieces are left out
sched.sanitize()
print(20 * '+', "before run")
sched.list(details=verbose)
print(20 * 'x')
if sched.orchestrate():
if verbose:
print(20 * '+', "after run")
sched.list()
print(20 * 'x')
print("image {} OK".format(self.to_image))
return True
else:
print("Something went wrong with image {}".format(self.to_image))
print(20 * '+', "after run - KO")
sched.debrief()
print(20 * 'x')
return False
def main() -> bool:
parser = ArgumentParser(formatter_class=ArgumentDefaultsHelpFormatter)
parser.add_argument("-U",
"--url",
default=default_topurl,
dest='topurl',
help="url to reach nbhosting server")
parser.add_argument(
"-i",
"--indices",
default=[0],
action=IntsRanges,
help="(cumulative) ranges of indices in the list of known notebooks"
" - run nbhtest with -l to see list")
parser.add_argument(
"-u",
"--users",
default=[1],
action=IntsRanges,
help=
"(cumulative) ranges of students indexes; e.g. -u 101-400 -u 501-600")
parser.add_argument("-b",
"--base",
default='student',
help="basename for students name")
parser.add_argument("-p",
"--period",
default=default_period,
type=float,
help="delay between 2 triggers of nbhtest")
parser.add_argument(
"-s",
"--sleep",
default=default_sleep_internal,
type=float,
help="delay in seconds to sleep between actions inside nbhtest")
parser.add_argument(
"-g",
"--go",
default=default_go_between_notebooks,
type=float,
help="go/wait duration between 2 consecutive notebooks")
parser.add_argument(
"-c",
"--cut",
default=False,
action='store_true',
help="""just load the urls, don't do any further processing""")
parser.add_argument("-w",
"--window",
default=default_window,
type=int,
help="window depth for spawning the nbhtest instances")
parser.add_argument("--idle", default=None, help="monitor idle setting")
parser.add_argument("-n", "--dry-run", action='store_true')
parser.add_argument(
"coursedirs",
default=[default_course_gitdir],
nargs='*',
help="""a list of git repos where to fetch notebooks""")
signature = "".join(sys.argv[1:])
args = parser.parse_args()
local = LocalNode(formatter=TerminalFormatter(
custom_format="%H-%M-%S:@line@", verbose=True))
if args.idle is not None:
hostname = urlparse(args.topurl).netloc
command = f"ssh root@{hostname} nbh test-set-monitor-idle {args.idle}"
os.system(command)
scheduler = Scheduler()
for user in args.users:
student_name = f"{args.base}-{user:04d}"
for coursedir in args.coursedirs:
command = (f"nbhtest.py -U {args.topurl} -u {student_name} "
f"-s {args.sleep} -g {args.go} ")
if args.cut:
command += "-c "
for index in args.indices:
command += f"{coursedir}:{index} "
command += " &"
if args.dry_run:
print("dry-run:", command)
else:
# schedule this command to run
_job = SshJob(scheduler=scheduler,
node=local,
commands=[command, f"sleep {args.period}"])
if args.dry_run:
return True
scheduler.jobs_window = args.window
overall = scheduler.orchestrate()
if not overall:
scheduler.debrief()
untagged = Path("artefacts")
tagged = Path(f"artefacts{signature}")
if tagged.exists():
print(f"NOT RENAMING because {tagged} exists; command to run is")
print(f"mv {untagged} {tagged}")
else:
print(f"renaming {untagged} into {tagged}")
untagged.rename(tagged)
print("nbhtests DONE")
return overall
