def crds(authenticated_user):
get_local_bucket(empty=True)
creds = StorageCreds(user=authenticated_user, service_url="storage")
drive = Drive(creds=creds, name="test_drive")
filepath = os.path.join(
os.path.dirname(__file__),
"../../../tests/data/proc_test_data/CRDS/bsd.picarro.1minute.248m.dat",
)
filemeta = drive.upload(filepath)
par = PAR(location=filemeta.location(), user=authenticated_user)
hugs = Service(service_url="hugs")
par_secret = hugs.encrypt_data(par.secret())
auth = Authorisation(resource="process", user=authenticated_user)
args = {
"authorisation": auth.to_data(),
"par": {
"data": par.to_data()
},
"par_secret": {
"data": par_secret
},
"data_type": "CRDS",
"source_name": "bsd.picarro.1minute.248m",
}
hugs.call_function(function="process", args=args)
def test_process_CRDS(authenticated_user, tempdir):
creds = StorageCreds(user=authenticated_user, service_url="storage")
drive = Drive(creds=creds, name="test_drive")
filepath = os.path.join(
os.path.dirname(__file__),
"../../../tests/data/proc_test_data/CRDS/bsd.picarro.1minute.248m.dat",
)
filemeta = drive.upload(filepath)
Path("/tmp/bsd.picarro.1minute.248m.dat").unlink(missing_ok=True)
par = PAR(location=filemeta.location(), user=authenticated_user)
hugs = Service(service_url="hugs")
par_secret = hugs.encrypt_data(par.secret())
auth = Authorisation(resource="process", user=authenticated_user)
args = {
"authorisation": auth.to_data(),
"par": {
"data": par.to_data()
},
"par_secret": {
"data": par_secret
},
"data_type": "CRDS",
"source_name": "bsd.picarro.1minute.248m",
}
response = hugs.call_function(function="process", args=args)
expected_keys = [
"bsd.picarro.1minute.248m_ch4",
"bsd.picarro.1minute.248m_co",
"bsd.picarro.1minute.248m_co2",
]
results = response["results"]["bsd.picarro.1minute.248m.dat"]
return False
assert sorted(results.keys()) == expected_keys
def test_chunking(authenticated_user, tempdir):
drive_name = "test_chunking"
creds = StorageCreds(user=authenticated_user, service_url="storage")
drive = Drive(name=drive_name, creds=creds)
uploader = drive.chunk_upload("test_chunking.py")
uploader.upload("This is some text\n")
uploader.upload("Here is")
uploader.upload(" some more!\n")
uploader.close()
filename = drive.download("test_chunking.py", dir=tempdir)
lines = open(filename).readlines()
assert(lines[0] == "This is some text\n")
assert(lines[1] == "Here is some more!\n")
downloader = drive.chunk_download("test_chunking.py", dir=tempdir)
filename = downloader.local_filename()
nchunks = 0
while downloader.download_next_chunk():
nchunks = nchunks + 1
if nchunks > 4:
assert(False)
assert(nchunks == 4) # extra chunk is empty to close
lines = open(filename).readlines()
assert(lines[0] == "This is some text\n")
assert(lines[1] == "Here is some more!\n")
def test_process_GC(authenticated_user, tempdir):
creds = StorageCreds(user=authenticated_user, service_url="storage")
drive = Drive(creds=creds, name="test_drive")
data_filepath = os.path.join(
os.path.dirname(__file__),
"../../../tests/data/proc_test_data/GC/capegrim-medusa.18.C",
)
precision_filepath = os.path.join(
os.path.dirname(__file__),
"../../../tests/data/proc_test_data/GC/capegrim-medusa.18.precisions.C",
)
Path("/tmp/capegrim-medusa.18.C").unlink(missing_ok=True)
Path("/tmp/capegrim-medusa.18.precisions.C").unlink(missing_ok=True)
data_meta = drive.upload(data_filepath)
precision_meta = drive.upload(precision_filepath)
data_par = PAR(location=data_meta.location(), user=authenticated_user)
precision_par = PAR(location=precision_meta.location(),
user=authenticated_user)
hugs = Service(service_url="hugs")
data_secret = hugs.encrypt_data(data_par.secret())
precision_secret = hugs.encrypt_data(precision_par.secret())
auth = Authorisation(resource="process", user=authenticated_user)
args = {
"authorisation": auth.to_data(),
"par": {
"data": data_par.to_data(),
"precision": precision_par.to_data()
},
"par_secret": {
"data": data_secret,
"precision": precision_secret
},
"data_type": "GCWERKS",
"source_name": "capegrim-medusa",
"site": "CGO",
"instrument": "medusa",
}
response = hugs.call_function(function="process", args=args)
result_keys = (sorted(
response["results"]["capegrim-medusa.18.C"].keys()))[:8]
expected_keys = [
"capegrim-medusa.18_C4F10",
"capegrim-medusa.18_C6F14",
"capegrim-medusa.18_CCl4",
"capegrim-medusa.18_CF4",
"capegrim-medusa.18_CFC-11",
"capegrim-medusa.18_CFC-112",
"capegrim-medusa.18_CFC-113",
"capegrim-medusa.18_CFC-114",
]
assert result_keys == expected_keys
def create_job(
self,
auth_user,
requirements,
key_password,
data_files,
hugs_url=None,
storage_url=None,
):
""" Create a job
Args:
auth_user (Acquire.User): Authenticated Acquire user
The following keys are required:
"hostname", "username", "name", "run_command", "partition", "n_nodes", "n_tasks_per_node",
"n_cpus_per_task", "memory_req", "job_duration"
where partition must be one of:
"cpu_test", "dcv", "gpu", "gpu_veryshort", "hmem", "serial", "test", "veryshort"
Example:
requirements = {"hostname": hostname, "username": username, "name": "test_job,
"n_nodes": 2,"n_tasks_per_node": 2,
"n_cpus_per_task": 2, "memory": "128G", ...}
requirements (dict): Dictionary containing job details and requested resources
key_password (str): Password for private key used to access the HPC
data_files (dict): Data file(s) to be uploaded to the cloud drive to
run the simulation. Simulation code files should be given in the "app" key and data
files in the "data" key
TODO - having to pass in a password and get it through to Paramiko seems
long winded, is there a better way to do this?
hugs_url (str): URL of HUGS service
storage_url (str): URL of storage service
Returns:
dict: Dictionary containing information regarding job running on resource
This will contain the PAR for access for data upload and download.
"""
from Acquire.Client import (
Drive,
Service,
PAR,
Authorisation,
StorageCreds,
Location,
ACLRule,
)
from Acquire.ObjectStore import create_uuid
import datetime
import os
if self._service is None:
raise PermissionError("Cannot use a null service")
if storage_url is None:
storage_url = self._service_url + "/storage"
if hugs_url is None:
hugs_url = self._service_url + "/hugs"
if not isinstance(data_files["app"], list):
data_files["app"] = [data_files["app"]]
try:
if not isinstance(data_files["data"], list):
data_files["data"] = [data_files["data"]]
except KeyError:
pass
# Get an authorisaton to pass to the service
hugs = Service(service_url=hugs_url)
# Credentials to create the cloud storage drive
creds = StorageCreds(user=auth_user, service_url=storage_url)
# Append a shortened UUID to the job name to ensure we don't multiple drives with the same name
short_uuid = create_uuid(short_uid=True)
job_name = requirements["name"]
job_name = f"{job_name.lower()}_{short_uuid}"
# Create a cloud drive for the input and output data to be written to
drive = Drive(creds=creds, name=job_name)
# Check the size of the files and if we want to use the chunk uploader
# Now we want to upload the files to the cloud drive we've created for this job
chunk_limit = 50 * 1024 * 1024
# Store the metadata for the uploaded files
uploaded_files = {"app": {}, "data": {}}
# These probably won't be very big so don't check their size
for f in data_files["app"]:
file_meta = drive.upload(f, dir="app")
uploaded_files["app"][f] = file_meta
# We might not have any data files to upload
try:
for f in data_files["data"]:
filesize = os.path.getsize(f)
if filesize < chunk_limit:
file_meta = drive.upload(f, dir="data")
else:
file_meta = drive.chunk_upload(f, dir="data")
uploaded_files["data"][f] = file_meta
except KeyError:
pass
auth = Authorisation(resource="job_runner", user=auth_user)
# Create a PAR with a long lifetime here and return a version to the user
# and another to the server to allow writing of result data
drive_guid = drive.metadata().guid()
location = Location(drive_guid=drive_guid)
# Read the duration from the requirements dictionary
# TODO - add in some reading of the duration
# try:
# duration = requirements["duration"]
# par_expiry = datetime.datetime
par_lifetime = datetime.datetime.now() + datetime.timedelta(days=1)
# Create an ACL rule for this PAR so we can read and write to it
aclrule = ACLRule.owner()
par = PAR(
location=location,
user=auth_user,
aclrule=aclrule,
expires_datetime=par_lifetime,
)
par_secret = par.secret()
encryped_par_secret = hugs.encrypt_data(par_secret)
# Encrypt the password we use to decrypt the private key used to access the HPC cluster
# TODO - is this a sensible way of doing this?
encrypted_password = hugs.encrypt_data(key_password)
par_data = par.to_data()
args = {}
args["authorisation"] = auth.to_data()
args["par"] = par_data
args["par_secret"] = encryped_par_secret
args["requirements"] = requirements
args["key_password"] = encrypted_password
function_response = self._service.call_function(function="job_runner",
args=args)
response = {}
response["function_response"] = function_response
response["par"] = par_data
response["par_secret"] = par_secret
response["upload_data"] = uploaded_files
return response
def test_drive_par(authenticated_user, tempdir):
drive_name = "test å∫ç∂ pars"
creds = StorageCreds(user=authenticated_user, service_url="storage")
drive = Drive(name=drive_name, creds=creds)
drive.upload(filename=__file__, uploaded_name="tmp_test.py")
downloaded_name = drive.download(filename="tmp_test.py", dir=tempdir)
assert (_same_file(__file__, downloaded_name))
drive_guid = drive.metadata().guid()
location = Location(drive_guid=drive_guid)
par = PAR(location=location,
user=authenticated_user,
aclrule=ACLRule.reader())
par_drive = par.resolve()
assert (par_drive.metadata().acl() == ACLRule.reader())
assert (par_drive.metadata().uid() == drive.metadata().uid())
files = par_drive.list_files()
assert (len(files) == 1)
assert (files[0].filename() == "tmp_test.py")
downloaded_name = files[0].open().download(dir=tempdir, force_par=True)
assert (_same_file(__file__, downloaded_name))
par2 = PAR(location=location,
user=authenticated_user,
aclrule=ACLRule.writer())
par_drive = par2.resolve()
assert (par_drive.metadata().acl() == ACLRule.writer())
assert (par_drive.metadata().uid() == drive.metadata().uid())
files = par_drive.list_files()
assert (len(files) == 1)
assert (files[0].filename() == "tmp_test.py")
par_drive.upload(filename=__file__, uploaded_name="tmp_test2.py")
files = par_drive.list_files()
assert (len(files) == 2)
f = {}
f[files[0].filename()] = files[0]
f[files[1].filename()] = files[1]
files = f
assert ("tmp_test.py" in files)
assert ("tmp_test2.py" in files)
downloaded_name = files["tmp_test2.py"].open().download(dir=tempdir)
assert (_same_file(__file__, downloaded_name))
par = PAR(location=files["tmp_test.py"].location(),
user=authenticated_user,
aclrule=ACLRule.reader())
par_file = par.resolve()
assert (par_file.metadata().acl() == ACLRule.reader())
downloaded_name = par_file.download(dir=tempdir)
assert (_same_file(__file__, downloaded_name))
with pytest.raises(PermissionError):
par_file.upload(__file__)
par = PAR(location=files["tmp_test.py"].location(),
user=authenticated_user,
aclrule=ACLRule.writer())
par_file = par.resolve()
assert (par_file.metadata().acl() == ACLRule.writer())
par_file.upload(__file__)
def process_files(
self,
user,
files,
data_type,
source_name=None,
overwrite=False,
hugs_url=None,
storage_url=None,
datasource=None,
site=None,
instrument=None,
):
""" Process the passed file(s)
Args:
user (User): Authenticated Acquire User
files (str, list): Path of files to be processed
data_type (str): Type of data to be processed (CRDS, GC etc)
hugs_url (str): URL of HUGS service. Currently used for testing
datasource (str): Datasource name or UUID
This may be removed in the future.
storage_url (str): URL of storage service. Currently used for testing
This may be removed in the future.
site (str, default=None): Name of site, three letter code or long name
instrument (str, default=None): If no instrument name is passed we will attempt
to find it from the filename.
Returns:
dict: UUIDs of Datasources storing data of processed files keyed by filename
"""
data_type = data_type.upper()
if self._service is None:
raise PermissionError("Cannot use a null service")
if not isinstance(files, list):
files = [files]
if data_type.upper() == "GC":
if not all(isinstance(item, tuple) for item in files):
return TypeError(
"If data type is GC, a list of tuples for data and precision filenames must be passed"
)
files = [(Path(f), Path(p)) for f, p in files]
else:
files = [Path(f) for f in files]
if storage_url is None:
storage_url = self._service_url + "/storage"
if hugs_url is None:
hugs_url = self._service_url + "/hugs"
# # Take the filename without the file extension
# source_name = [os.path.splitext((filepath.name).split("/")[-1])[0] for filepath in files]
hugs = Service(service_url=hugs_url)
creds = StorageCreds(user=user, service_url=storage_url)
drive = Drive(creds=creds, name="test_drive")
auth = Authorisation(resource="process", user=user)
# Here we'll need special cases for different data types. As GC requires
# both the data file and precision data and they need to be kept together
# for use in processing.
# We can maybe reconsider the way this is done if there ends up being a lot of test
# cases and this gets a bit clunky
results = {}
for file in files:
if data_type == "GC":
if source_name is None:
source_name = file[0].stem
if site is None:
site = source_name.split(".")[0]
if "-" in site and data_type == "GC":
site = site.split("-")[0]
filemeta = drive.upload(file[0])
par = PAR(location=filemeta.location(), user=user)
par_secret = hugs.encrypt_data(par.secret())
prec_meta = drive.upload(file[1])
prec_par = PAR(location=prec_meta.location(), user=user)
prec_par_secret = hugs.encrypt_data(prec_par.secret())
args = {
"authorisation": auth.to_data(),
"par": {"data": par.to_data(), "precision": prec_par.to_data()},
"par_secret": {"data": par_secret, "precision": prec_par_secret},
"data_type": data_type,
"datasource": datasource,
"source_name": source_name,
"overwrite": overwrite,
"site": site,
"instrument": instrument,
}
else:
filemeta = drive.upload(file)
par = PAR(location=filemeta.location(), user=user)
par_secret = hugs.encrypt_data(par.secret())
args = {
"authorisation": auth.to_data(),
"par": {"data": par.to_data()},
"par_secret": {"data": par_secret},
"data_type": data_type,
"datasource": datasource,
"source_name": source_name,
"overwrite": overwrite,
}
# If we try to upload many files we don't want it to fail if a single
# file contains overlapping data
try:
response = self._service.call_function(function="process", args=args)
results.update(response["results"])
except ValueError as err:
results[file.name] = err
return results
def test_drives(authenticated_user, tempdir):
creds = StorageCreds(user=authenticated_user, service_url="storage")
nstart = len(Drive.list_toplevel_drives(creds=creds))
drive_name = "test å∫ç∂ something"
drive = Drive(name=drive_name, creds=creds, autocreate=True)
assert (drive.metadata().name() == drive_name)
assert (drive.metadata().acl().is_owner())
drive2_name = "test/this/is/a/../../dir"
drive2 = Drive(name=drive2_name, creds=creds)
drives = Drive.list_toplevel_drives(creds=creds)
assert (len(drives) == nstart + 2)
drives = drive2.list_drives()
assert (len(drives) == 0)
drives = drive.list_drives()
assert (len(drives) == 0)
filename = __file__
files = drive.list_files()
assert (len(files) == 0)
filemeta = drive.upload(filename=filename)
assert (filemeta.is_complete())
assert (filemeta.acl().is_owner())
assert (filemeta.acl().is_readable())
assert (filemeta.acl().is_writeable())
assert (filemeta.uploaded_by() == authenticated_user.guid())
assert (filemeta.uploaded_when() is not None)
upload_datetime = filemeta.uploaded_when()
(_, filename) = os.path.split(filename)
assert (filemeta.filename() == filename)
files = drive.list_files()
assert (len(files) == 1)
assert (files[0].filename() == filemeta.filename())
assert (not files[0].is_complete())
files = drive.list_files(include_metadata=True)
assert (len(files) == 1)
assert (files[0].filename() == filemeta.filename())
assert (files[0].is_complete())
assert (files[0].uid() == filemeta.uid())
assert (files[0].filesize() == filemeta.filesize())
assert (files[0].checksum() == filemeta.checksum())
assert (files[0].compression_type() == filemeta.compression_type())
assert (files[0].uploaded_by() == filemeta.uploaded_by())
assert (files[0].uploaded_when() == filemeta.uploaded_when())
assert (files[0].acl().is_owner())
assert (files[0].uploaded_by() == authenticated_user.guid())
assert (files[0].uploaded_when() == upload_datetime)
f = files[0].open()
filename = f.download(dir=tempdir)
# make sure that the two files are identical
with open(filename, "rb") as FILE:
data1 = FILE.read()
# remove this tmp file
os.unlink(filename)
with open(__file__, "rb") as FILE:
data2 = FILE.read()
assert (data1 == data2)
assert (files[0].uid() == filemeta.uid())
assert (files[0].filesize() == filemeta.filesize())
assert (files[0].checksum() == filemeta.checksum())
assert (files[0].compression_type() == filemeta.compression_type())
assert (files[0].uploaded_by() == filemeta.uploaded_by())
assert (files[0].uploaded_when() == filemeta.uploaded_when())
assert (files[0].acl().is_owner())
assert (files[0].uploaded_by() == authenticated_user.guid())
assert (files[0].uploaded_when() == upload_datetime)
versions = f.list_versions()
assert (len(versions) == 1)
assert (versions[0].filename() == filemeta.filename())
assert (versions[0].uploaded_when() == filemeta.uploaded_when())
new_filemeta = drive.upload(filename=__file__, force_par=True)
versions = f.list_versions()
assert (len(versions) == 2)
filename = new_filemeta.open().download(dir=tempdir)
# make sure that the two files are identical
with open(filename, "rb") as FILE:
data1 = FILE.read()
# remove this tmp file
os.unlink(filename)
with open(__file__, "rb") as FILE:
data2 = FILE.read()
assert (data1 == data2)
# should be in upload order
assert (versions[0].uid() == filemeta.uid())
assert (versions[1].uid() == new_filemeta.uid())
filename = new_filemeta.open().download(dir=tempdir, force_par=True)
# make sure that the two files are identical
with open(filename, "rb") as FILE:
data1 = FILE.read()
# remove this tmp file
os.unlink(filename)
assert (data1 == data2)
# try to upload a file with path to the drive
filemeta = drive.upload(filename=__file__,
uploaded_name="/test/one/../two/test.py")
assert (filemeta.filename() == "test/two/test.py")
# cannot create a new Drive with non-owner ACLs
with pytest.raises(PermissionError):
drive = Drive(name="broken_acl",
creds=creds,
aclrules=ACLRules.owner("12345@z0-z0"))
drive = Drive(name="working_acl",
creds=creds,
aclrules=ACLRules.owner(authenticated_user.guid()))
def test_run_calc(aaai_services, authenticated_user):
# create and register the cluster on which this job will take place...
cluster = Cluster.create(service_url="compute",
user=aaai_services["compute"]["user"])
user = authenticated_user
assert (user.is_logged_in())
# ensure that the main account has money
deposit(user,
100.0,
"Adding money to the account",
accounting_url="accounting")
# get a handle to the financial account used to pay for the job
account = Account(user=user,
account_name="deposits",
accounting_url="accounting")
assert (account.balance() >= 100.0)
# Upload a directory that will contain all of the input
creds = StorageCreds(user=user, service_url="storage")
drive = Drive(name="sim", creds=creds, autocreate=True)
uploaded = drive.upload(_testdata())
location = uploaded.location()
print(drive.list_files(dir="example_sim/input"))
print(location)
# create a request for a job to be run using:
# 1. 'image_name' as the specified container image for the software
# 2. 'location' as the location containing all input files
# authorise it using the authenticated user (who may be different to the
# user who pays for the job - hence the need for a different
# authorisation for the request and for the cheque)
r = RunRequest(image="docker://test_image:latest", input=location)
# now write a cheque which will provide authorisation to spend money from
# this account to pay for this request. This will be written to the access
# service to give it the authority to create a transation in the account.
# This cheque authorises only a single transaction, performable only
# by the service whose canonical URL is supplied, and the access service
# should check that the requested resource signature matches that
# authorised by the cheque
cheque = Cheque.write(account=account,
recipient_url="access",
resource=r.fingerprint(),
max_spend=50.0)
func = "run_calculation"
args = {}
args["request"] = r.to_data()
args["authorisation"] = Authorisation(user=user,
resource=r.fingerprint()).to_data()
args["cheque"] = cheque.to_data()
access_service = Service("access")
result = access_service.call_function(func, args)
print(result)
pending_uids = cluster.get_pending_job_uids()
print(pending_uids)
for uid in pending_uids:
job = cluster.submit_job(uid)
print(job)
pending_uids = cluster.get_pending_job_uids()
print(pending_uids)