def stop_calc(conn, project_codename):
"""
Tell the calculation to stop, running a specific script on the worker
:param conn: The ssh connection to the main worker
:type conn: ssh.SshConnection
:param project_codename: The uuid of the project
:type project_codename: str
"""
calc_dir = util.path_join(api_util.WORKER_WORK_PATH, "ZephyTOOLS",
"PROJECTS_CFD", project_codename, "CALC")
if not conn.folder_exists(calc_dir):
log.debug("calc folder " + calc_dir +
" doesn't exists yet, skipping...")
return True
_, out, _ = conn.run(
["find", calc_dir, "-mindepth", "1", "-maxdepth", "1", "-type", "d"])
out = out.strip()
if not out or "\n" in out: # No results or more than one result
log.warning("Unable to get the calculation output folder")
return
calc_dir = out.rstrip("/")
stopper_command = util.path_join(api_util.WORKER_WORK_PATH, "ZephyTOOLS",
"APPLI", "TMP", "CFD_CALC_STOP.py")
conn.run(["python", stopper_command, calc_dir])
def check_status(self):
if not self._running_worker.is_debug:
if models.jobs.is_shutdown_disabled(self._job_id):
self._running_worker.disable_shutdown()
if self.is_canceled():
self._status = models.jobs.JOB_STATUS_CANCELED
raise api_util.ToolchainCanceled()
status_file = util.path_join(api_util.WORKER_OUTPUT_PATH, "task_end.txt")
code, out, err = self.conn.run(["cat", status_file], can_fail=True)
if code == 0:
task_status = out.strip()
if task_status == "success":
self._status = models.jobs.JOB_STATUS_FINISHED
elif task_status == "cancel":
self._status = models.jobs.JOB_STATUS_CANCELED
elif task_status == "error":
self._status = models.jobs.JOB_STATUS_KILLED
else:
raise RuntimeError("Unknown worker task result: " + repr(task_status))
return self._status
# Ensure the proc is still running
if not self._proc.is_running():
self._status = models.jobs.JOB_STATUS_KILLED
# Load progress
exit_code, out, _ = self.conn.run(["cat", api_util.WORKER_WORK_PATH + "/progress.txt"], can_fail=True)
if exit_code == 0 and type_util.ll_float(out.strip()):
models.jobs.set_job_progress(self._job_id, max(0.0, min(1.0, float(out))))
return self._status
def stop_and_wait(self):
self.conn.run("echo '1' > '" + util.path_join(api_util.WORKER_INPUT_PATH, "output_fetched")+"'", shell=True)
try:
proc_util.wait_for_proc_and_streams(self._proc, 2)
except util.TimeoutError:
pass # Machine stopped
self._proc = None
def __enter__(self):
proc = None
try:
log_info_file = util.path_join(api_util.WORKER_INPUT_PATH, "log_info.json")
log_info = {
"jobid": self._job_id,
"job_type": self._command,
"api_name": self._api_name,
"server_name": self._server_name,
"instance": self._running_worker.worker_ids[0],
"provider": self._running_worker.provider_name
}
with file_util.temp_file(json.dumps(log_info)) as tmp_filepath:
self.conn.send_file(tmp_filepath, api_util.WORKER_INPUT_PATH + "/log_info.json")
self.conn.run(["chmod", "a+r", log_info_file])
proc = self.conn.run_async(["python", api_util.WORKER_RUNNER_PATH])
# Send the task parameter to the worker
task_params_file = os.path.join(api_util.WORKER_INPUT_PATH, "task_params.json")
task_params = {
"jobid": self._job_id,
"project_uid": self._project_uid,
"toolchain": self._command,
"params": json.dumps(self._params),
"shutdown": "0" if self._running_worker.is_debug else "1"
}
with file_util.temp_file(json.dumps(task_params)) as tmp_filepath:
self.conn.send_file(tmp_filepath, task_params_file)
self.conn.run(["chmod", "a+r", task_params_file])
self._proc = proc
except error_util.abort_errors:
with error_util.before_raising():
log.info("Signal received, stopping process")
if proc:
proc_util.ensure_stop_proc(proc, 2)
except error_util.all_errors:
with error_util.before_raising():
try:
if proc:
proc_util.ensure_kill_proc(proc)
except error_util.abort_errors:
pass
except error_util.all_errors as e:
logging.getLogger("aziugo").exception(e)
return self
from lib import type_util
import provider
import storages
API_PATH = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", ".."))
PRICE_PRECISION = 100000000 # 10^8 (could be a problem after 92 billions dollar/zephycoins)
CURRENCY_DOLLAR = "dollar"
CURRENCY_YUAN = "yuan"
CURRENCY_EURO = "euro"
OPENFOAM_DONATION_RATIO = 0.05
# Worker path definitions
WORKER_HOME = "/home/aziugo"
WORKER_SCRIPTS_PATH = util.path_join(WORKER_HOME, "worker_scripts")
WORKER_INPUT_PATH = util.path_join(WORKER_SCRIPTS_PATH, "inputs")
WORKER_OUTPUT_PATH = util.path_join(WORKER_SCRIPTS_PATH, "outputs")
WORKER_RUNNER_PATH = util.path_join(WORKER_SCRIPTS_PATH, "aziugo_start.py")
WORKER_LAUNCHER_PATH = util.path_join(WORKER_SCRIPTS_PATH, "python_venv.sh")
WORKER_WORK_PATH = util.path_join(WORKER_SCRIPTS_PATH, "workdir")
log = logging.getLogger("aziugo")
class NoMoreCredits(RuntimeError):
pass
class ToolchainError(RuntimeError):
pass
def calculate(api_name, server_name, job, project, mesh, calculation,
calc_param_file, provider_name, machine, nbr_machines,
split_results, client_login, client_ip, api_version):
"""
Launch the machine(s), send the files, start the worker script, wait for progress and results and saving results
:param api_name: The name of the api
:type api_name: str
:param server_name: The name of the server (ex: apidev.zephycloud.com)
:type server_name: str
:param job: The job information
:type job: dict[str, any]
:param project: The main project
:type project: dict[str, any]
:param mesh: The mesh used for this calculation
:type mesh: dict[str, any]
:param calculation: The calculation to launch
:type calculation: dict[str, any]
:param calc_param_file: The main job parameter file
:type calc_param_file: str
:param provider_name: The name of the provider
:type provider_name: str
:param machine: The type of machine to launch
:type machine: str
:param nbr_machines: The number of machines to run
:type nbr_machines: int
:param split_results: Do you want the result file to be splitted ?
:type split_results: bool
:param client_login: The login of the job owner
:type client_login: str
:param client_ip: The client ip address of the http request string this job
:type client_ip: str
:param api_version: The version of the http api where the user ask to launch this job
:type api_version: str
"""
job_id = int(job['id'])
nbr_machines = int(nbr_machines)
models.jobs.set_job_status(job_id, models.jobs.JOB_STATUS_RUNNING)
project_codename = project['uid']
user_id = project["user_id"]
calc_id = calculation['id']
tmp_folder = api_util.get_conf().get("general", "tmp_folder")
provider = api_util.get_provider(provider_name)
storage = api_util.get_storage(project['storage'])
tags = {
'operation': "calc",
'job_id': str(job_id),
'server': server_name,
'api': api_name,
'api_version': api_version,
'client': client_login,
'client_ip': client_ip,
'debug': DO_NOT_KILL_INSTANCES,
'trusted': IS_TOOLCHAIN_SECURED
}
models.users.charge_user_fix_price(user_id, job_id,
"Calculation storage cost")
result_name = project_codename + "-calc-" + str(job_id)
result_file = cmd_util.ResultFile(project_codename, result_name + ".zip")
internal_file = cmd_util.ResultFile(project_codename,
result_name + "_workfiles.zip")
if split_results:
iterations_file = cmd_util.ResultFile(project_codename,
result_name + "_iterations.zip")
reduce_file = cmd_util.ResultFile(project_codename,
result_name + "_reduce.zip")
# Uploading file on cloud storage
log.info("Uploading param file to storage")
models.calc.save_calc_param_file(user_id, project_codename,
calculation['name'], calc_param_file)
try:
# Creating worker
with cmd_util.using_workers(
api_name,
provider,
job_id,
machine,
nbr_machines,
tags,
debug_keep_instances_alive=DO_NOT_KILL_INSTANCES) as workers:
# Launch main script
with cmd_util.TaskProcess(job_id, job["project_uid"], "calc",
workers, [split_results]) as task_proc:
conn = workers.ssh_connection
# Charge user
end_time = models.users.charge_user_computing(
user_id, job_id, "Cloud computation cost")
if models.users.get_credit(user_id) <= 0:
raise api_util.NoMoreCredits()
log.info("Sending project files on worker")
worker_in_storage = storages.SshStorage(
conn, api_util.WORKER_INPUT_PATH, IS_TOOLCHAIN_SECURED)
cmd_util.copy_project_file(
user_id,
project_codename,
storage,
worker_in_storage,
"project_file.zip",
tmp_folder,
key=models.projects.PROJECT_FILE_RAW)
cmd_util.copy_project_file(
user_id,
project_codename,
storage,
worker_in_storage,
"anal.zip",
tmp_folder,
key=models.projects.PROJECT_FILE_ANALYSED)
cmd_util.copy_project_file(user_id,
project_codename,
storage,
worker_in_storage,
"mesh.zip",
tmp_folder,
file_id=mesh['result_file_id'])
worker_in_storage.upload_file(calc_param_file,
"calc_params.zip")
os.remove(calc_param_file)
log.info("Project files sent to the worker")
# Tell the script to start
log.info("Starting the computation")
task_proc.start()
last_fetched_progress_time = datetime.datetime.utcfromtimestamp(
0)
is_stopped = False
while True:
task_status = task_proc.check_status()
# Charge if we need
if datetime.datetime.utcnow() > end_time:
end_time = models.users.charge_user_computing(
user_id, job_id, "Cloud computation cost")
if models.users.get_credit(user_id) <= 0:
models.jobs.save_job_text(job_id, "No more credit")
raise api_util.NoMoreCredits()
if task_status != models.jobs.JOB_STATUS_RUNNING:
log.info("Computation finished with status: " +
models.jobs.job_status_to_str(task_status))
break
if (datetime.datetime.utcnow() - last_fetched_progress_time
).seconds > STATUS_FETCHING_DELAY:
fetch_progress(conn, user_id, project_codename,
calculation['name'], calculation['id'],
storage, tmp_folder)
last_fetched_progress_time = datetime.datetime.utcnow()
if not is_stopped:
calculation = models.calc.get_calc(
user_id, project['uid'], calculation['id'])
if not calculation:
raise api_util.ToolchainError("Calculation " +
str(calc_id) +
" disappeared")
if calculation['status'] == models.calc.STATUS_STOPPED:
log.info("Stopping computation")
stop_calc(conn, project_codename)
is_stopped = True
time.sleep(1)
# Checking if the machine is still here
if not conn.ping():
models.jobs.save_job_text(job_id,
"Worker instance disappeared")
raise api_util.ToolchainError(
"Worker instance disappeared")
# Fetching computed data
log.info("Saving results")
worker_out_storage = storages.SshStorage(
conn, api_util.WORKER_OUTPUT_PATH, IS_TOOLCHAIN_SECURED)
log_file = util.path_join(api_util.WORKER_OUTPUT_PATH,
"worker.log")
if conn.file_exists(log_file):
with file_util.temp_filename(dir=tmp_folder) as tmp:
conn.get_file(log_file, tmp)
models.jobs.save_job_log(job_id, tmp)
else:
log.warning("No worker log file")
if not result_file.exists(worker_out_storage):
log.error("Unable to find file " + str(result_file) +
" on worker")
raise api_util.ToolchainError(
"Task failed, no result file")
result_file.save_on_storage(worker_out_storage, storage,
tmp_folder)
if split_results:
if not iterations_file.exists(worker_out_storage):
log.error("Unable to find file " +
str(iterations_file) + " on worker")
raise api_util.ToolchainError(
"Task failed, no result file")
iterations_file.save_on_storage(worker_out_storage,
storage, tmp_folder)
if not reduce_file.exists(worker_out_storage):
log.error("Unable to find file " + str(reduce_file) +
" on worker")
raise api_util.ToolchainError(
"Task failed, no result file")
reduce_file.save_on_storage(worker_out_storage, storage,
tmp_folder)
if internal_file.exists(worker_out_storage):
internal_file.save_on_storage(worker_out_storage, storage,
tmp_folder)
else:
log.warning("No internal files found on server")
fetch_progress(conn, user_id, project_codename,
calculation['name'], calculation['id'], storage,
tmp_folder)
log.info("Computation result fetched")
# Signaling all output was fetched
task_proc.stop_and_wait()
# Charge again if required
if datetime.datetime.utcnow() > end_time:
models.users.charge_user_computing(project["user_id"], job_id,
"Cloud computation cost")
# Uploading file on cloud storage
result_file.save_in_database(user_id)
internal_file_id = internal_file.save_in_database(
user_id) if internal_file.saved else None
if split_results:
iterations_file.save_in_database(user_id)
reduce_file.save_in_database(user_id)
models.calc.save_result_files(user_id, project_codename,
calculation['name'],
result_file.file_id,
iterations_file.file_id,
reduce_file.file_id,
internal_file_id)
else:
models.calc.save_result_files(user_id, project_codename,
calculation['name'],
result_file.file_id, None, None,
internal_file_id)
except error_util.all_errors:
with error_util.before_raising():
if REMOVE_RESULTS_ON_ERROR:
internal_file.delete_from_distant(storage)
result_file.delete_from_distant(storage)
if split_results:
iterations_file.delete_from_distant(storage)
reduce_file.delete_from_distant(storage)
log.info("Results saved")
def fetch_progress(conn, user_id, project_codename, calc_name, calc_id,
storage, tmp_folder):
"""
Check if a progress file has been created on the main worker and save it if it exists
:param conn: The ssh connection to the main worker
:type conn: ssh.SshConnection
:param user_id: The id of the job owner
:type user_id: int
:param project_codename: The project uuid
:type project_codename: str
:param calc_name: The name of the calculation
:type calc_name: str
:param calc_id: The id of the calculation
:type calc_id: int
:param storage: The storage of the project
:type storage: core.ssh.Storage
:param tmp_folder: A temporary folder to use
:type tmp_folder: str
:return: True if success, False if no file is found or a failure happens
:rtype: bool
"""
status_file_name = project_codename + "_calc_" + calc_name + "_status.zip"
status_file = cmd_util.ResultFile(project_codename, status_file_name)
old_status_file = None
try:
calc_dir = util.path_join(api_util.WORKER_WORK_PATH, "ZephyTOOLS",
"PROJECTS_CFD", project_codename, "CALC")
if not conn.folder_exists(calc_dir):
log.debug("calc folder " + calc_dir +
" doesn't exists yet, skipping...")
return True
_, out, _ = conn.run([
"find", calc_dir, "-mindepth", "1", "-maxdepth", "1", "-type", "d"
])
out = out.strip()
if not out or "\n" in out: # No results or more than one result
log.warning("Unable to get the calculation output folder")
return
calc_dir = out.rstrip("/")
zipper_command = util.path_join(api_util.WORKER_WORK_PATH,
"ZephyTOOLS", "APPLI", "TMP",
"CFD_CALC_ZIP_STATUS.py")
old_status_file = models.calc.get_calc_status_file(
user_id, project_codename, calc_id)
status_file_path = util.path_join(api_util.WORKER_OUTPUT_PATH,
status_file_name)
conn.run(
["python", zipper_command, "-i", calc_dir, "-o", status_file_path])
worker_out_storage = storages.SshStorage(conn,
api_util.WORKER_OUTPUT_PATH,
IS_TOOLCHAIN_SECURED)
if not status_file.exists(worker_out_storage):
log.warning(
"Unable to get calculation status file: file not found")
return False
status_file.save_on_storage(worker_out_storage, storage, tmp_folder)
file_id = status_file.save_in_database(user_id)
models.calc.save_status_file(user_id, project_codename, calc_id,
file_id)
except error_util.all_errors as e:
with error_util.saved_stack() as error_stack:
status_file.delete_from_distant(storage)
if error_util.is_abort(e):
error_stack.reraise()
else:
error_util.log_error(log, e)
return False
if old_status_file:
models.projects.remove_file_from_project(user_id, project_codename,
old_status_file['id'])
return True
def start(self):
self.conn.run(["rm", "-f", util.path_join(api_util.WORKER_OUTPUT_PATH, "task_end.txt")])
self.conn.run("echo '1' > '" + util.path_join(api_util.WORKER_INPUT_PATH, "start_task")+"'", shell=True)
self._status = models.jobs.JOB_STATUS_RUNNING
def disable_shutdown(self):
self._debug = True
self._provider.tag_workers(self._workers, {'debug': True})
flag_file = util.path_join(api_util.WORKER_INPUT_PATH, "disable_shutdown")
self._conn.run("echo '1' > '" + flag_file + "'", shell=True)
def analyse(api_name, server_name, job, project, storage_name, project_file,
provider_name, machine, nbr_machines, client_login, client_ip,
api_version):
"""
Launch the machine(s), send the files, start the worker script, wait for progress and results and saving results
:param api_name: The name of the api
:type api_name: str
:param server_name: The name of the server (ex: apidev.zephycloud.com)
:type server_name: str
:param job: The job information
:type job: dict[str, any]
:param project: The main project
:type project: dict[str, any]
:param storage_name: The name of the storage where the project will be located
:type storage_name: str
:param project_file: The raw project file to analyse
:type project_file: str
:param provider_name: The name of the provider
:type provider_name: str
:param machine: The type of machine to launch
:type machine: str
:param nbr_machines: The number of machines to run
:type nbr_machines: int
:param client_login: The login of the job owner
:type client_login: str
:param client_ip: The client ip address of the http request string this job
:type client_ip: str
:param api_version: The version of the http api where the user ask to launch this job
:type api_version: str
"""
job_id = int(job['id'])
project_codename = job["project_uid"]
analyzed_filename = job["project_uid"] + "-anal-" + str(job_id) + ".zip"
user_id = project["user_id"]
provider = api_util.get_provider(provider_name)
storage = api_util.get_storage(storage_name)
tmp_folder = api_util.get_conf().get("general", "tmp_folder")
tags = {
'operation': "anal",
'job_id': str(job_id),
'server': server_name,
'api': api_name,
'api_version': api_version,
'client': client_login,
'client_ip': client_ip,
'trusted': IS_TOOLCHAIN_SECURED
}
# Uploading file on cloud storage
log.info("Uploading project file to storage")
models.projects.append_file_to_project(
user_id,
job["project_uid"],
project_file,
"project_" + job["project_uid"] + ".zip",
key=models.projects.PROJECT_FILE_RAW,
overwrite=True)
log.info("Project file uploaded")
models.users.charge_user_fix_price(user_id, job_id, "Project storage cost")
analyzed_file = cmd_util.ResultFile(project_codename, analyzed_filename)
try:
# Creating worker
with cmd_util.using_workers(
api_name,
provider,
job_id,
machine,
nbr_machines,
tags,
debug_keep_instances_alive=DO_NOT_KILL_INSTANCES) as workers:
with cmd_util.TaskProcess(job_id, job["project_uid"], "anal",
workers) as task_proc:
conn = workers.ssh_connection
# Charge user
end_time = models.users.charge_user_computing(
user_id, job_id, "Cloud computation cost")
if models.users.get_credit(user_id) <= 0:
raise api_util.NoMoreCredits()
log.info("Sending project files on worker")
conn.send_file(
project_file,
util.path_join(api_util.WORKER_INPUT_PATH,
"project_file.zip"))
os.remove(project_file)
log.info("Project files sent to the worker")
# Tell the script to start
log.info("Starting the computation")
task_proc.start()
while True:
task_status = task_proc.check_status()
# Charge if we need
if datetime.datetime.utcnow() > end_time:
end_time = models.users.charge_user_computing(
user_id, job_id, "Cloud computation cost")
if models.users.get_credit(user_id) <= 0:
models.jobs.save_job_text(job_id, "No more credit")
raise api_util.NoMoreCredits()
if task_status != models.jobs.JOB_STATUS_RUNNING:
log.info("Computation finished with status: " +
models.jobs.job_status_to_str(task_status))
break
time.sleep(5)
# Checking if the machine is still here
if not conn.ping():
models.jobs.save_job_text(job_id,
"Worker instance disappeared")
raise api_util.ToolchainError(
"Worker instance disappeared")
# Fetching computed data
log.info("Fetching results")
worker_out_storage = storages.SshStorage(
conn, api_util.WORKER_OUTPUT_PATH, IS_TOOLCHAIN_SECURED)
log_file = util.path_join(api_util.WORKER_OUTPUT_PATH,
"worker.log")
if conn.file_exists(log_file):
with file_util.temp_filename(dir=tmp_folder) as tmp:
conn.get_file(log_file, tmp)
models.jobs.save_job_log(job_id, tmp)
else:
log.warning("No worker log file")
if not analyzed_file.exists(worker_out_storage):
log.error("Unable to find file " + str(analyzed_file) +
" on worker")
raise api_util.ToolchainError(
"Task failed, no result file")
analyzed_file.save_on_storage(worker_out_storage, storage,
tmp_folder)
log.info("Computation result fetched")
# Signaling all output was fetched
task_proc.stop_and_wait()
# Charge again if required
if datetime.datetime.utcnow() > end_time:
models.users.charge_user_computing(project["user_id"], job_id,
"Cloud computation cost")
# Uploading file on cloud storage
analyzed_file.save_in_database(
user_id, key=models.projects.PROJECT_FILE_ANALYSED)
except error_util.all_errors:
with error_util.before_raising():
if REMOVE_RESULTS_ON_ERROR:
analyzed_file.delete_from_distant(storage)
log.info("Result saved")