def setup(self,master,hosts):
#
# SETUP MASTER
#
if self.master:
banner(f"Setting up master {master}")
self.run_script(name="spark.prereqs", hosts=master)
self.run_script(name="spark.download.spark", hosts=master)
self.run_script(name="spark.install", hosts=master)
self.run_script(name="spark.bashrc.master", hosts=master)
#
# SETUP WORKER
#
if self.workers:
# Copy setup files to each worker and execute the shell program in parallel
workers = ', '.join(hosts)
banner(f"Setup Workers: {workers}")
command = "sh ~/spark-setup-worker.sh"
jobSet = JobSet("spark_worker_install", executor=JobSet.ssh)
for host in hosts:
banner(f"Setting up worker {host}")
command1 = f"scp /home/pi/cm/cloudmesh-pi-cluster/cloudmesh/pi/cluster/spark/bin/spark-setup-worker.sh pi@{host}:"
print(command1)
os.system(command1)
command2 = f"scp ~/sparkout.tgz pi@{host}:"
print(command2)
os.system(command2)
command3 = f"scp /home/pi/cm/cloudmesh-pi-cluster/cloudmesh/pi/cluster/spark/bin/spark-bashrc.txt pi@{host}:"
print(command3)
os.system(command3)
jobSet.add({"name": host, "host": host, "command": command})
self.update_slaves(host)
jobSet.run(parallel=len(hosts))
jobSet.Print()
return
def run(self, beginLine=None, endLine=None):
# Prepare script
self.script = textwrap.dedent(str(self.script))
lines = self.script.splitlines()
# Truncate lines if begin and/or line parameters are set
if beginLine is not None and endLine is not None:
lines = Shell.find_lines_between(lines, beginLine, endLine)
elif beginLine is not None and endLine is None:
lines = Shell.find_lines_from(lines, beginLine)
elif beginLine is None and endLine is not None:
lines = Shell.find_lines_to(lines, endLine)
# Loop over each line
for line in lines:
stripped = line.strip()
# Check if line is a comment
if stripped.startswith("#") or stripped == "":
pass
else:
# Create jobSet per line
job = JobSet(self.name, executor=self.executor)
if "# tag:" in line:
line, tag = line.split("# tag:", 1)
tag = tag.strip()
line = line.strip()
# Execute jobSet for each host in parallel
for host in self.hosts:
job.add({"name": host, "host": host, "command": line});
job.run(parallel=len(self.hosts))
job.Print()
def uninstall(self, hosts):
job_set = JobSet("mongo_install", executor=JobSet.ssh)
command = """
sudo apt-get -y remove mongodb
sudo apt-get -y remove --purge mongodb
sudo apt-get autoremove
python3 -m pip uninstall pymongo
"""
for host in hosts:
job_set.add({"name": host, "host": host, "command": command})
job_set.run(parallel=len(hosts))
job_set.Print()
banner("MongoDB Removed Succesfully")
return
def stop(self, master, workers):
hosts = Parameter.expand(workers)
command = "sudo service mongodb stop"
# Stop mongo
if master is not None:
os.system(command)
if hosts is not None:
job_set = JobSet("mongo_stop", executor=JobSet.ssh)
for host in hosts:
job_set.add({"name": host, "host": host, "command": command})
job_set.run(parallel=len(hosts))
# job_set.Print()
banner("MongoDB service stopped succesfully")
return 1
def run(self,
script=None,
name="script",
host=None,
executor=JobSet.ssh,
**kwargs):
# Prepare parameters
if script is None:
Console.error("The script is not defined, found None as content")
return
host = host or os.uname()[1]
if kwargs:
parameters = dotdict(**kwargs)
else:
parameters = dotdict({})
parameters.host = host
# Prepare script
self.script = textwrap.dedent(str(script))
self.script = self.script.format(**kwargs)
lines = self.script.splitlines()
# Add script to jobset and run
jobs = JobSet("onejob", executor=executor)
counter = 1
for line in lines:
stripped = line.strip()
if stripped.startswith("#") or stripped == "":
pass
else:
tag = counter
if "# tag:" in line:
line, tag = line.split("# tag:", 1)
tag = tag.strip()
line = line.strip()
jobs.add({
"script": name,
"name": tag,
"tag": tag,
"line": counter,
"host": host,
"counter": counter,
"command": line
})
counter = counter + 1
jobs.run(parallel=1)
self.job = jobs.job
return self.job
def ssh_keygen(hosts=None,
filename="~/.ssh/id_rsa",
username=None,
processors=3,
dryrun=False,
verbose=True):
"""
generates the keys on the specified hosts.
this fonction does not work well as it still will aski if we overwrite.
:param hosts:
:param filename:
:param username:
:param output:
:param dryrun:
:param verbose:
:return:
"""
hosts = Parameter.expand(hosts)
command = f'ssh-keygen -q -N "" -f {filename} <<< y'
jobSet = JobSet("ssh_keygen", executor=JobSet.ssh)
for host in hosts:
jobSet.add({"name": host, "host": host, "command": command})
jobSet.run(parallel=len(hosts))
#result_keys = Host.ssh(hosts=hosts,
# command=command,
# username=username,
# dryrun=dryrun,
# processors=processors,
# executor=os.system)
#result_keys = Host.ssh(hosts=hosts,
# processors=processors,
# command='cat .ssh/id_rsa.pub',
# username=username)
jobSet.Print()
def delete(self, master=None, hosts=None):
# Delete master node
# TODO - k3s does not allow you to delete it's parent node
#if master is not None:
# banner(f"Deleting Master Node: {master}")
#
# command = Installer.oneline(f"""
# sudo kubectl delete {master}
# """)
# jobSet = JobSet("kubernetes_master_delete", executor=JobSet.ssh)
# jobSet.add({"name": self.hostname, "host": master, "command": command})
# jobSet.run()
# Uninstall workers
if hosts is not None:
workers = ', '.join(hosts)
banner(f"Deleting Worker Node(s): {workers}")
jobSet = JobSet("kubernetes_worker_delete", executor=JobSet.ssh)
for host in hosts:
command = Installer.oneline(f"""
sudo kubectl drain {host} --ignore-daemonsets --delete-local-data;
sudo kubectl delete node {host}
""")
jobSet.add({
"name": self.hostname,
"host": self.hostname,
"command": command
})
jobSet.run(parallel=len(hosts))
print(
"Workers:",
Printer.write(jobSet.array(),
order=[
"name", "command", "status", "stdout",
"returncode"
]))
def uninstall(self, master, hosts):
#
# Uninstall MASTER
#
if self.master:
banner(f"Uninstall Master: {master}")
self.run_script(name="spark.uninstall.master", hosts=master)
#
# Uninstall WORKER(S)
#
if self.workers:
banner(f"Uninstall Workers: {hosts}")
command7 = "sh ~/spark-uninstall-worker.sh"
jobSet = JobSet("spark_worker_uninstall", executor=JobSet.ssh)
for host in hosts:
jobSet.add({"name": host, "host": host, "command": command7})
jobSet.run(parallel=len(hosts))
#jobSet.Print()
banner("Successfully uninstalled workers")
return
def install(self, hosts):
job_set = JobSet("mongo_install", executor=JobSet.ssh)
command = """
sudo apt update
sudo apt -y upgrade
sudo apt -y install mongodb
sudo apt-get -y install python3-pip
python3 -m pip install pymongo
mkdir -p /home/pi/data/db
"""
# Install mongodb on all Pis
for host in hosts:
job_set.add({"name": host, "host": host, "command": command})
job_set.run(parallel=len(hosts))
job_set.Print()
# Copy config files to all hosts
# Have not used JobSet here. Copying serially...
for host in hosts:
command = f"scp /home/pi/cm/cloudmesh-pi-cluster/cloudmesh/pi/cluster/mongo/bin/local_setup.cfg pi@{host}:/home/pi/mongodb.conf"
os.system(command)
banner("MongoDB Setup and Configuration Complete")
def enable_containers(self, filename="/boot/cmdline.txt", hosts=None):
line = "cgroup_enable=cpuset cgroup_memory=1 cgroup_enable=memory"
warning = "Your master already has containers enabled"
self.add_to_file(filename, line, warning)
if hosts is not None:
# Create tmp file on master with enable_containers line
source = "~/container_tmp.txt"
command = f'echo "{line}" >> {source}'
os.system(command)
# Copy over temporary file with container line
for host in hosts:
command = "scp {0} pi@{1}:~/".format(source, host)
os.system(command)
# Check if workers already have line and if not, append to /boot/cmdline.txt
tmp_cmdline = "~/cmdline.txt"
command = f"""
if grep -q "{line}" '/boot/cmdline.txt'
then
rm {source};
else
cp /boot/cmdline.txt {tmp_cmdline};
cat {source} >> {tmp_cmdline};
sudo cp {tmp_cmdline} {filename}; rm {tmp_cmdline} {source};
fi"""
jobSet = JobSet("kubernetes_worker_enable_containers",
executor=JobSet.ssh)
for host in hosts:
jobSet.add({"name": host, "host": host, "command": command})
jobSet.run()
#jobSet.Print()
# Delete tmp file on master
command = f"rm {source}"
os.system(command)
def uninstall(self, master=None, hosts=None):
# Uninstall master
if master is not None:
banner(f"Uninstalling Master {master}")
command = "/usr/local/bin/k3s-uninstall.sh"
jobSet = JobSet("kubernetes_master_uninstall", executor=JobSet.ssh)
jobSet.add({"name": master, "host": master, "command": command})
jobSet.run()
jobSet.Print()
# Uninstall workers
if hosts is not None:
workers = ', '.join(hosts)
banner(f"Uninstalling Workers: {workers}")
#command = "/usr/local/bin/k3s-agent-uninstall.sh"
command = "/usr/local/bin/k3s-uninstall.sh"
jobSet = JobSet("kubernetes_worker_uninstall", executor=JobSet.ssh)
for host in hosts:
jobSet.add({"name": host, "host": host, "command": command})
jobSet.run(parallel=len(hosts))
jobSet.Print()
self.delete(hosts=hosts)
def join(self, master=None, hosts=None):
if hosts is not None and master is not None:
banner(f"Get Join Token From {master}")
command = "sudo cat /var/lib/rancher/k3s/server/node-token"
jobSet = JobSet("kubernetes_token_retrieval", executor=JobSet.ssh)
jobSet.add({"name": master, "host": master, "command": command})
jobSet.run()
token = jobSet.array()[0]['stdout'].decode('UTF-8')
# TODO - Currently get ip address from eth0 instead of using hostname
# because worker does not know master's host name
ip = self.get_master_ip_address('eth0')
command = f"""
sudo k3s agent --server https://{ip}:{self.port}
--token {token}"""
jobSet = JobSet("kubernetes_worker_join", executor=JobSet.ssh)
for host in hosts:
jobSet.add({"name": host, "host": host, "command": command})
jobSet.run(parallel=len(hosts))
jobSet.Print()
self.view()
def install(self, master=None, hosts=None, step=None):
# Setup containers on master
if master is None and hosts:
Console.error("You must specify a master to set up nodes")
raise ValueError
if step is not None:
if step in 'enable_containers':
self.enable_containers(hosts=hosts)
# Install K3S on the master
if master is not None:
if type(master) != list:
master = Parameter.expand(master)
#
# TODO - bug I should be able to run this even if I am not on master
#
banner(f"Setup Master: {master[0]}")
command = Installer.oneline(f"""
curl -sfL https://get.k3s.io | sh -
""")
jobSet = JobSet("kubernetes_master_install", executor=JobSet.ssh)
jobSet.add({
"name": self.hostname,
"host": master[0],
"command": command
})
jobSet.run()
result_stdout = jobSet.array()[0]['stdout'].decode('UTF-8')
if "No change detected" in result_stdout:
print()
Console.info("Service already running")
# Setup workers and join to cluster
#
# TODO - bug I should be able to run this even if I am not on master
#
if hosts is not None:
if master is not None:
banner(f"Get Join Token From {master[0]}")
command = "sudo cat /var/lib/rancher/k3s/server/node-token"
jobSet = JobSet("kubernetes_token_retrieval",
executor=JobSet.ssh)
jobSet.add({
"name": master[0],
"host": master[0],
"command": command
})
jobSet.run()
token = jobSet.array()[0]['stdout'].decode('UTF-8')
# Install kubernetes to workers
workers = ', '.join(hosts)
banner(f"Setup Workers: {workers}")
command = "curl -sfL http://get.k3s.io | sh -"
#worker_install = Host.ssh(hosts=hosts, command=command, executor=os.system)
jobSet = JobSet("kubernetes_worker_install",
executor=JobSet.ssh)
for host in hosts:
jobSet.add({
"name": host,
"host": host,
"command": command
})
jobSet.run(parallel=len(hosts))
jobSet.Print()
# Join workers to master's cluster
# TODO - Currently get ip address from eth0 instead of using hostname
# because worker does not know master's host name
ip = self.get_master_ip_address('eth0')
command = f"""sudo k3s agent --server https://{ip}:{self.port}
--token {token}"""
# TODO - This currently does not work, command runs fine but
# "k3s agent" having trouble creating node.
jobSet = JobSet("kubernetes_worker_join", executor=JobSet.ssh)
for host in hosts:
jobSet.add({
"name": host,
"host": host,
"command": command
})
jobSet.run(parallel=len(hosts))
jobSet.Print()
else:
Console.warning(
"You must have the master parameter set to burn workers")
# Print created cluster
self.view()
def do_host(self, args, arguments):
"""
::
Usage:
host scp NAMES SOURCE DESTINATION [--dryrun]
host ssh NAMES COMMAND [--dryrun] [--output=FORMAT]
host config NAMES [IPS] [--user=USER] [--key=PUBLIC]
host check NAMES [--user=USER] [--key=PUBLIC]
host key create NAMES [--user=USER] [--dryrun] [--output=FORMAT]
host key list NAMES [--output=FORMAT]
host key gather NAMES [--authorized_keys] [FILE]
host key scatter USERNAME NAMES FILE
This command does some useful things.
Arguments:
FILE a file name
Options:
--dryrun shows what would be done but does not execute
--output=FORMAT the format of the output
Description:
host scp NAMES SOURCE DESTINATION
TBD
host ssh NAMES COMMAND
runs the command on all specified hosts
Example:
ssh red[01-10] \"uname -a\"
host key create NAMES
create a ~/.ssh/id_rsa and id_rsa.pub on all hosts specified
Example:
ssh key create "red[01-10]"
host key list NAMES
list all id_rsa.pub keys from all hosts specifed
Example:
ssh key list red[01-10]
host key gather HOSTS FILE
gathers all keys from file FILE including the one from localhost.
ssh key gather "red[01-10]" keys.txt
host key scatter USERNAME HOSTS FILE
copies all keys from file FILE to authorized_keys on all hosts,
but also makes sure that the users ~/.ssh/id_rsa.pub key is in
the file.
1) adds ~/.id_rsa.pub to the FILE only if its not already in it
2) removes all duplicated keys
Example:
ssh key scatter pi "red[01-10]" ~/keys.txt
host key scp NAMES FILE
copies all keys from file FILE to authorized_keys on all hosts
but also makes sure that the users ~/.ssh/id_rsa.pub key is in
the file and removes duplicates, e.g. it calls fix before upload
Example:
ssh key list red[01-10] > pubkeys.txt
ssh key scp red[01-10] pubkeys.txt
host config NAMES IPS [--user=USER] [--key=PUBLIC]
generates an ssh config file tempalte that can be added to your
.ssh/config file
Example:
cms host config "red,red[01-03]" "198.168.1.[1-4]" --user=pi
host check NAMES [--user=USER] [--key=PUBLIC]
This command is used to test if you can login to the specified
hosts. It executes the hostname command and compares it.
It provides a table with a sucess column
cms host check "red,red[01-03]"
+-------+---------+--------+
| host | success | stdout |
+-------+---------+--------+
| red | True | red |
| red01 | True | red01 |
| red02 | True | red02 |
| red03 | True | red03 |
+-------+---------+--------+
"""
def _print(results):
arguments.output = arguments.output or 'table'
if arguments.output == 'table':
print(
Printer.write(results, order=['host', 'success',
'stdout']))
else:
pprint(results)
map_parameters(arguments, 'dryrun', 'output', 'user')
dryrun = arguments.dryrun
if dryrun:
VERBOSE(arguments)
if arguments.scp and not arguments.key:
destinations = Parameter.expand(arguments.DESTINATION)
source = arguments.SOURCE
results_key = Host.scp(source, destinations, output="lines")
elif arguments.ssh:
names = Parameter.expand(arguments.NAMES)
# print (names)
results = Host.ssh(hosts=names, command=arguments.COMMAND)
_print(results)
elif arguments.key and arguments.create:
Host.ssh_keygen(hosts=arguments.NAMES,
username=arguments.user,
dryrun=dryrun)
#_print(results)
elif arguments.key and arguments.list:
names = Parameter.expand(arguments.NAMES)
jobSet = JobSet("key_list", executor=JobSet.ssh)
command = "cat .ssh/id_rsa.pub"
for host in names:
jobSet.add({"name": host, "host": host, "command": command})
jobSet.run(parallel=len(names))
#results = Host.ssh(hosts=names,
# command='cat .ssh/id_rsa.pub',
# username=arguments.user)
jobSet.Print()
elif arguments.key and arguments.gather:
output = Host.gather_keys(username=arguments.user,
hosts=arguments.NAMES,
filename="~/.ssh/id_rsa.pub",
key="~/.ssh/id_rsa",
processors=3,
dryrun=False)
if arguments.FILE:
filename = path_expand(arguments.FILE)
directory = os.path.dirname(filename)
if directory:
Shell.mkdir(directory)
with open(filename, "w") as f:
f.write(output)
else:
print(output)
elif arguments.key and arguments.scatter:
names = arguments.NAMES
file = arguments.get("FILE")
if not os.path.isfile(file):
Console.error("The file does not exist")
return ""
Host.put(hosts=names,
source=file,
username=arguments.USERNAME,
destination=".ssh/authorized_keys")
#_print(result)
elif arguments.config:
key = arguments.key or "~/.ssh/id_rsa.pub"
result = Host.config(hosts=arguments.NAMES,
ips=arguments.IPS,
username=arguments.user,
key=key)
print(result)
elif arguments.check:
key = arguments.key or "~/.ssh/id_rsa.pub"
result = Host.check(hosts=arguments.NAMES,
username=arguments.user,
key=key)
for entry in result:
entry['success'] = entry['stdout'] == entry['host']
_print(result)
return ""
def gather_keys(username=None,
hosts=None,
filename="~/.ssh/id_rsa.pub",
key="~/.ssh/id_rsa",
processors=3,
dryrun=False):
"""
returns in a list the keys of the specified hosts
:param username:
:param hosts:
:param filename:
:param key:
:param dryrun:
:return:
"""
names = Parameter.expand(hosts)
jobSet = JobSet("ssh_keygen", executor=JobSet.ssh)
command = "cat .ssh/id_rsa.pub"
for host in names:
jobSet.add({"name": host, "host": host, "command": command})
jobSet.run(parallel=len(hosts))
results_key = []
#jobSet.Print()
for key in jobSet.array():
stdout = key['stdout'].decode('UTF-8')
if "Command could not run" not in stdout:
results_key.append(stdout)
#results_key = Host.ssh(hosts=names,
# command='cat .ssh/id_rsa.pub',
# username=username,
# verbose=False)
#results_authorized = Host.ssh(hosts=names,
# command='cat .ssh/id_rsa.pub',
# username=username,
# verbose=False)
filename = path_expand(filename)
localkey = {
'host': "localhost",
'command': [''],
'execute': "",
'stdout': readfile(filename).strip(),
'stderr': None,
'returncode': True,
'success': True,
'date': DateTime.now()
}
if results_key is None: # and results_authorized is None:
return ""
# geting the output and also removing duplicates
output = [localkey['stdout']] + \
results_key
output = '\n'.join(output) + "\n"
return output
def put(hosts=None,
source=None,
destination=None,
username=None,
key="~/.ssh/id_rsa.pub",
shell=False,
processors=3,
dryrun=False,
verbose=False):
"""
:param command: the command to be executed
:param hosts: a list of hosts to be checked
:param username: the usernames for the hosts
:param key: the key for logging in
:param processors: the number of parallel checks
:return: list of dicts representing the ping result
"""
hosts = Parameter.expand(hosts)
key = path_expand(key)
# Copy over your file to temporary destination on each host
jobSet = JobSet("put_tmp_file", executor=JobSet.execute)
master = os.uname()[1]
randomInt = random.randint(1000, 2000)
tmp_destination = "~/.cloudmesh/tmp_file-" + str(randomInt)
for host in hosts:
command = f"""ssh {username}@{host} "mkdir -p ~/.cloudmesh/";
scp {source} {username}@{host}:{tmp_destination}"""
jobSet.add({"name": host, "host": master, "command": command})
jobSet.run(parallel=len(hosts))
jobSet.Print()
# Overwrite desintation file with your tempoary file
jobSet = JobSet("put_overwrite", executor=JobSet.ssh)
command = f"sudo cp {tmp_destination} {destination}; rm -f {tmp_destination}"
for host in hosts:
jobSet.add({"name": host, "host": host, "command": command})
jobSet.run(parallel=len(hosts))
jobSet.Print()