def inspect(self, name):
"""Inspect a local image in the database, which typically includes the
basic fields in the model.
"""
container = self.get(name)
if container is not None:
collection = container.collection.name
fields = container.__dict__.copy()
fields["collection"] = collection
fields["metrics"] = json.loads(fields["metrics"])
del fields["_sa_instance_state"]
fields["created_at"] = str(fields["created_at"])
print(json.dumps(fields, indent=4, sort_keys=True))
return fields
bot.exit("image {} was not found in the database".format(name))
def _update_secrets(self):
'''update secrets will take a secrets credential file
either located at .sregistry or the environment variable
SREGISTRY_CLIENT_SECRETS and update the current client
secrets as well as the associated API base. For the case of
using Docker Hub, if we find a .docker secrets file, we update
from there.
'''
bot.debug('Creating aws client...')
try:
from awscli.clidriver import create_clidriver
except:
bot.exit('Please install pip install sregistry[aws]')
driver = create_clidriver()
self.aws = driver.session.create_client('ecr')
def stream_response(self, response, stream_to=None, show_progress=True):
"""
stream response is one level higher up than stream, starting with a
response object and then performing the stream without making the
requests.get. The expectation is that the request was successful
(status code 20*).
show_progress: boolean to show progress bar
Parameters
==========
response: a response that is ready to be iterated over to download in
streamed chunks
stream_to: the file to stream to
"""
if response.status_code == 200:
if show_progress is False:
bot.quiet = True
# Keep user updated with Progress Bar, if not quiet
content_size = None
if "Content-Length" in response.headers:
progress = 0
content_size = int(response.headers["Content-Length"])
bot.show_progress(progress, content_size, length=35)
chunk_size = 1 << 20
with open(stream_to, "wb") as filey:
for chunk in response.iter_content(chunk_size=chunk_size):
filey.write(chunk)
if content_size is not None:
progress += chunk_size
bot.show_progress(
iteration=progress,
total=content_size,
length=35,
carriage_return=False,
)
# Newline to finish download
sys.stdout.write("\n")
return stream_to
bot.exit("Problem with stream, response %s" % (response.status_code))
def authorize(self, names, payload=None, request_type="push"):
"""Authorize a client based on encrypting the payload with the client
token, which should be matched on the receiving server"""
self.username = self._get_and_update_setting("SREGISTRY_REGISTRY_USERNAME")
if self.username is None:
# backwards compatibility
self.username = self._get_and_update_setting("username")
if self.username is None:
bot.exit(
"Failed to authorize: please set SREGISTRY_REGISTRY_USERNAME to an appropriate value"
)
else:
self._update_setting("SREGISTRY_REGISTRY_USERNAME", self.username)
self.token = self._get_and_update_setting("SREGISTRY_REGISTRY_TOKEN")
if self.token is None:
# backwards compatibility
self.token = self._get_and_update_setting("token")
if self.token is None:
bot.exit(
"Failed to authorize: please set SREGISTRY_REGISTRY_TOKEN to an appropriate value"
)
else:
self._update_setting("SREGISTRY_REGISTRY_TOKEN", self.token)
# Use the payload to generate a digest push|collection|name|tag|user
timestamp = generate_timestamp()
credential = generate_credential(self.username)
credential = "%s/%s/%s" % (request_type, credential, timestamp)
if payload is None:
payload = "%s|%s|%s|%s|%s|" % (
request_type,
names["collection"],
timestamp,
names["image"],
names["tag"],
)
signature = generate_signature(payload, self.token)
return "SREGISTRY-HMAC-SHA256 Credential=%s,Signature=%s" % (credential,
signature)
def main(args, parser, extra):
"""the images entrypoint is intended to list images locally in the user
database, optionally taking one or more query string to subset the
search
"""
from sregistry.main import get_client
cli = get_client(quiet=args.quiet)
# If the client doesn't have the command, exit
if not hasattr(cli, "images"):
bot.exit("listing images requires using the sqlite database.")
for query in args.query:
if query in ["", "*"]:
query = None
cli.images(query=query)
def main(args, parser, extra):
'''sharing an image means sending a remote share from an image you
control to a contact, usually an email.
'''
from sregistry.main import get_client
image = args.image
# Detect any uri, and refresh client if necessary
cli = get_client(image, quiet=args.quiet)
cli.announce(args.command)
# If the client doesn't have the command, exit
if not hasattr(cli, 'share'):
msg = "share is not implemented for %s. Why don't you add it?"
bot.exit(msg % cli.client_name)
cli.share(image, share_to=args.share_to)
def _client_tagged(self, tags):
'''ensure that the client name is included in a list of tags. This is
important for matching builders to the correct client. We exit
on fail.
Parameters
==========
tags: a list of tags to look for client name in
'''
# We must match the client to a tag
name = self.client_name.lower()
tags = [t.lower() for t in tags]
if name not in tags:
bot.exit('%s not found in %s, must match!' % (name, tags))
def _update_secrets(self):
"""update secrets will look for a dropbox token in the environment at
SREGISTRY_DROPBOX_TOKEN and if found, create a client. If not,
an error message is returned and the client exits.
"""
# Retrieve the user token. Exit if not found
token = self._required_get_and_update("SREGISTRY_DROPBOX_TOKEN")
# Create the dropbox client
self.dbx = Dropbox(token)
# Verify that the account is valid
try:
self.account = self.dbx.users_get_current_account()
except:
bot.exit("Account invalid. Exiting.")
def _get_bucket(self):
"""get a bucket based on a bucket name. If it doesn't exist, create it.
"""
# Case 1: The bucket already exists
try:
self._bucket = self._bucket_service.get_bucket(self._bucket_name)
# Case 2: The bucket needs to be created
except google.cloud.exceptions.NotFound:
self._bucket = self._bucket_service.create_bucket(self._bucket_name)
# Case 3: The bucket name is already taken
except:
bot.exit("Cannot get or create %s" % self._bucket_name)
return self._bucket
def main(args, parser, extra):
from sregistry.main import get_client
image = args.image
cli = get_client(image, quiet=args.quiet)
cli.announce(args.command)
if not hasattr(cli, 'rm'):
msg = "remove is not implemented for %s. Why don't you add it?"
bot.exit(msg % cli.client_name)
result = cli.rm(image)
if result is None:
bot.exit("No {} record found in the database".format(image))
def run_compute_build(cli, args):
"""a compute based build is the oldest versions of build - here we bring
up our own instance, and then provide control to it. The helper
functions below (kill, instances, templates) support this version.
"""
# Does the user want to save the image?
command = args.commands.pop(0)
# Option 1: The user wants to kill an instance
if command == "kill":
kill(args)
# Option 2: Just list running instances
elif command == "instances":
instances(args)
# Option 3: The user wants to list templates
elif "template" in command:
templates(args)
# Option 4: View a specific or latest log
elif command == "logs":
list_logs(args)
# Option 3: The user is providing a Github repo!
recipe = "Singularity"
if "github" in command:
# One argument indicates a recipe
if len(args.commands) == 1:
recipe = args.commands.pop(0)
else:
# If a command is provided, but not a Github repo
bot.exit("%s is not a recognized option." % command)
# Does the user want to specify a name for the collection?
name = args.name
# No image is needed, we are creating in the cloud
return cli.build(repo=command,
name=name,
recipe=recipe,
preview=args.preview)
def stream(
self,
url,
headers=None,
stream_to=None,
retry=True,
default_headers=True,
show_progress=True,
):
"""
stream is a get that will stream to file_name. This stream is intended
to take a url and (optionally) a set of headers and file to stream to,
and will generate a response with requests.get.
Parameters
==========
url: the url to do a requests.get to
headers: any updated headers to use for the requets
stream_to: the file to stream to
show_progress: boolean to show progress bar
retry: should the client retry? (intended for use after token refresh)
by default we retry once after token refresh, then fail.
"""
bot.debug("GET %s" % url)
# Ensure headers are present, update if not
if headers is None:
if self.headers is None:
self._reset_headers()
headers = self.headers.copy()
response = requests.get(url, headers=headers, verify=self._verify(), stream=True)
# Deal with token if necessary
if response.status_code == 401 and retry is True:
if hasattr(self, "_update_token"):
self._update_token(response)
return self.stream(
url, headers, stream_to, retry=False, show_progress=show_progress
)
if response.status_code == 200:
return self._stream(response, stream_to=stream_to, show_progress=show_progress)
bot.exit("Problem with stream, response %s" % (response.status_code))
def main(args, parser, extra):
from sregistry.main import get_client
for query in args.query:
original = query
query = remove_uri(query)
if query in ['', '*']:
query = None
try:
cli = get_client(original, quiet=args.quiet)
cli.announce(args.command)
cli.search(query, args=args)
except NotImplementedError:
msg = "search is not implemented for %s. Why don't you add it?"
bot.exit(msg % cli.client_name)
def main(args, parser, extra):
'''the list command corresponds with listing images for an external
resource. This is different from listing images that are local to the
database, which should be done with "images"
'''
from sregistry.main import get_client
cli = get_client(quiet=args.quiet)
# If the client doesn't have the command, exit
if not hasattr(cli, 'ls'):
msg = "list is not implemented for %s. Why don't you add it?"
bot.exit(msg % cli.client_name)
for query in args.query:
if query in ['', '*']:
query = None
cli.ls(query=query)
def _update_secrets(self):
"""The user is required to have an application secrets file in his
or her environment. The client exists with error
if the variable isn't found.
"""
env = "SREGISTRY_GOOGLE_DRIVE_CREDENTIALS"
self._secrets = self._get_and_update_setting(env)
self._base = self._get_and_update_setting(
"SREGISTRY_GOOGLE_DRIVE_ROOT")
if self._base is None:
self._base = "sregistry"
if self._secrets is None:
bot.error("You must export %s to use Google Drive client" % env)
bot.exit(
"https://singularityhub.github.io/sregistry-cli/client-google-drive"
)
def run_registry_build(cli, args, extra):
"""a registry build pushes a recipe file to Singularity Registry Server,
or given that a GitHub Url is provided, we build from there. For more
regular building, the user is suggested to directly connect the
repository to Singularity Registry server. This can serve as a one time
build.
"""
# The uri can also contain github, which indicates a Github build
if args.name is None:
bot.exit("Please provide a container identifier with --name")
recipe = args.commands.pop(0)
response = cli.build(name=args.name, recipe=recipe, extra=extra)
# Print output to the console
if response is not None:
print_output(response, args.outfile)
return response
def main(args, parser, extra):
from sregistry.main import get_client
# Does the user have a valid image?
image = args.image
if not os.path.exists(image):
bot.exit("%s does not exist" % image)
# Authenticate
cli = get_client(args.name, quiet=args.quiet)
cli.announce(args.command)
# If the client doesn't have the command, exit
if not hasattr(cli, "push"):
msg = "push is not implemented for %s. Why don't you add it?"
bot.exit(msg % cli.client_name)
cli.push(path=image, name=args.name, tag=args.tag)
def get_file_hash(image_path, algorithm='sha256'):
'''return an md5 hash of the file based on a criteria level. This
is intended to give the file a reasonable version.
Parameters
==========
image_path: full path to the singularity image
'''
try:
hasher = getattr(hashlib, algorithm)()
except AttributeError:
bot.error("%s is an invalid algorithm.")
bot.exit(' '.join(hashlib.algorithms_guaranteed))
with open(image_path, "rb") as f:
for chunk in iter(lambda: f.read(4096), b""):
hasher.update(chunk)
return hasher.hexdigest()
def init_db(self, db_path):
'''initialize the database, with the default database path or custom of
the format sqlite:////home/<username>/sregistry.db
The custom path can be set with the environment var SREGISTRY_DATABASE
when a user creates the client, we must initialize this db
the database should use the .singularity cache folder to cache
layers and images, and .singularity/sregistry.db as a database
'''
# Database Setup, use default if uri not provided
self.database = 'sqlite:///%s' % db_path
self.storage = SREGISTRY_STORAGE
# If the path isn't defined, cut out early
if not db_path:
return
# Ensure that the parent_folder exists)
parent_folder = os.path.dirname(db_path)
# Case 1: Does not exist
if not os.path.exists(parent_folder):
bot.exit("Database location {} does not exist.".format(parent_folder))
# Case 2: Insufficient permission for write
if not os.access(parent_folder, os.W_OK):
bot.exit(
"Insufficient permission to write to {}".format(parent_folder))
bot.debug("Database located at %s" % self.database)
self.engine = create_engine(self.database, convert_unicode=True)
self.session = scoped_session(
sessionmaker(autocommit=False, autoflush=False, bind=self.engine))
Base.query = self.session.query_property()
# import all modules here that might define models so that
# they will be registered properly on the metadata. Otherwise
# you will have to import them first before calling init_db()
Base.metadata.create_all(bind=self.engine)
self.Base = Base
def get_digests(self, repo_name, tag):
"""
return a list of layers from a manifest.
The function is intended to work with both version
1 and 2 of the schema. All layers (including redundant)
are returned. By default, we try version 2 first,
then fall back to version 1.
For version 1 manifests: extraction is reversed
Parameters
==========
manifest: the manifest to read_layers from
"""
if not hasattr(self, "manifest"):
bot.exit("Please retrieve manifest for the image first.")
# version 2 manifest here!
return self.manifest["layers"]
def get_bucket(self):
'''given a bucket name and a client that is initialized, get or
create the bucket.
'''
for attr in ['bucket_name', 's3']:
if not hasattr(self, attr):
bot.exit('client is missing attribute %s' % (attr))
# See if the bucket is already existing
self.bucket = None
for bucket in self.s3.buckets.all():
if bucket.name == self.bucket_name:
self.bucket = bucket
# If the bucket doesn't exist, create it
if self.bucket is None:
self.bucket = self.s3.create_bucket(Bucket=self.bucket_name)
bot.info('Created bucket %s' % self.bucket.name)
return self.bucket
def get_config(self, key="Entrypoint", delim=None):
"""get_config returns a particular key (default is Entrypoint)
from a VERSION 1 manifest obtained with get_manifest.
Parameters
==========
key: the key to return from the manifest config
delim: Given a list, the delim to use to join the entries.
Default is newline
"""
if not hasattr(self, "manifests"):
bot.exit("Please retrieve manifests for an image first.")
cmd = None
# If we didn't find the config value in version 2
if cmd is None and "config" in self.manifests:
# First try, version 2.0 manifest config has upper level config
manifest = self.manifests["config"]
if "config" in manifest:
if key in manifest["config"]:
cmd = manifest["config"][key]
# Second try, config manifest (not from verison 2.0 schema blob)
if cmd is None and "history" in manifest:
for entry in manifest["history"]:
if "v1Compatibility" in entry:
entry = json.loads(entry["v1Compatibility"])
if "config" in entry:
if key in entry["config"]:
cmd = entry["config"][key]
# Standard is to include commands like ['/bin/sh']
if isinstance(cmd, list):
if delim is not None:
cmd = delim.join(cmd)
bot.verbose("Found Docker config (%s) %s" % (key, cmd))
return cmd
def add(
self,
image_path=None,
image_uri=None,
image_name=None,
url=None,
metadata=None,
save=True,
copy=False,
):
"""dummy add simple returns an object that mimics a database entry, so the
calling function (in push or pull) can interact with it equally. Most
variables (other than image_path) are not used."""
# We can only save if the image is provided
if image_path is not None:
if not os.path.exists(image_path):
bot.exit("Cannot find %s" % image_path)
if image_uri is None:
bot.exit("You must provide an image uri <collection>/<namespace>")
names = parse_image_name(remove_uri(image_uri))
bot.debug("Added %s to filesystem" % names["uri"])
# Create a dummy container on the fly
class DummyContainer:
def __init__(self, image_path, client_name, url, names):
self.image = image_path
self.client = client_name
self.url = url
self.name = names["image"]
self.tag = names["tag"]
self.uri = names["uri"]
container = DummyContainer(image_path, self.client_name, url, names)
bot.info("[container][add] %s" % names["uri"])
return container
def get_blob_location(response, bucket):
"""return a relative path for a blob based on finding the build step,
and the container built from it.
Parameters
==========
response: the response from client._build_status(build_id)
bucket: the name of the build bucket.
"""
# Find the build step, it uses singularityware as a builder
build_step = [
x for x in response["steps"]
if x["name"] == "gcr.io/cloud-builders/gsutil" and x["args"][0] == "cp"
]
# This is fragile, but we have to get the container name
if len(build_step) > 0:
location = build_step[0]["args"][-1]
return re.sub("(gs://%s|%s)" % (bucket, bucket), "",
location).strip("/")
bot.exit("Cannot find build step with image name.")
def main(args, parser, extra):
from sregistry.main import get_client
image = args.image
name = args.name
# Customize client based on uri
cli = get_client(image, quiet=args.quiet)
cli.announce(args.command)
# Does the user want to save the image?
do_save = True
if args.nocache is True or not hasattr(cli, "storage"):
do_save = False
# If the client doesn't have the command, exit
if not hasattr(cli, "pull"):
msg = "pull is not implemented for %s. Why don't you add it?"
bot.exit(msg % cli.client_name)
cli.pull(images=image, file_name=name, force=args.force, save=do_save)
def get_blob_location(response, bucket):
'''return a relative path for a blob based on finding the build step,
and the container built from it.
Parameters
==========
response: the response from client._build_status(build_id)
bucket: the name of the build bucket.
'''
# Find the build step, it uses singularityware as a builder
build_step = [
x for x in response['steps']
if x['name'] == 'gcr.io/cloud-builders/gsutil' and x['args'][0] == 'cp'
]
# This is fragile, but we have to get the container name
if len(build_step) > 0:
location = build_step[0]['args'][-1]
return re.sub('(gs://%s|%s)' % (bucket, bucket), '',
location).strip('/')
bot.exit("Cannot find build step with image name.")
def update_token(self):
"""update_token uses HTTP basic authentication to get a token for
Docker registry API V2 operations. We get here if a 401 is
returned for a request.
Parameters
==========
response: the http request response to parse for the challenge.
https://docs.docker.com/registry/spec/auth/token/
"""
# Add Amazon headers
tokens = self.aws.get_authorization_token()
token = tokens["authorizationData"][0]["authorizationToken"]
try:
token = {"Authorization": "Basic %s" % token}
self.headers.update(token)
except:
bot.exit("Error getting token.")
def get_manifest(self, repo_name, tag):
"""return the image manifest via the aws client, saved in self.manifest
"""
image = None
repo = self.aws.describe_images(repositoryName=repo_name)
if "imageDetails" in repo:
for contender in repo.get("imageDetails"):
if tag in contender["imageTags"]:
image = contender
break
# if the image isn't found, we need to exit
if image is None:
bot.exit("Cannot find %s:%s, is the uri correct?" % (repo_name, tag))
digest = image["imageDigest"]
digests = self.aws.batch_get_image(
repositoryName=repo_name, imageIds=[{"imageDigest": digest, "imageTag": tag}]
)
self.manifest = json.loads(digests["images"][0]["imageManifest"])
return self.manifest
def load_build_config(self, config=None):
"""load a google compute config, meaning that we have the following cases:
1. the user has not provided a config file directly, we look in env.
2. the environment is not set, so we use a reasonable default
3. if the final string is not found as a file, we look for it in library
4. we load the library name, or the user file, else error
Parameters
==========
config: the config file the user has provided, or the library URI
"""
# If the config is already a dictionary, it's loaded
if isinstance(config, dict):
bot.debug("Config is already loaded.")
return config
# if the config is not defined, look in environment, then choose a default
if config is None:
config = self._get_and_update_setting(
"SREGISTRY_COMPUTE_CONFIG", "google/compute/ubuntu/securebuild-2.4.3"
)
# If the config is a file, we read it
elif os.path.exists(config):
return read_json(config)
# otherwise, try to look it up in library
configs = self._load_templates(config)
if configs is None:
bot.exit("%s is not a valid config. %s" % config)
bot.info("Found config %s in library!" % config)
config = configs[0]
return config
def _extract_tar(archive, output_folder):
'''use blob2oci to handle whiteout files for extraction. Credit for this
script goes to docker2oci by Olivier Freyermouth, and see script
folder for license.
Parameters
==========
archive: the archive to extract
output_folder the output folder (sandbox) to extract to
'''
from .terminal import (run_command, which)
result = which('blob2oci')
if result['return_code'] != 0:
bot.exit('Cannot find blob2oci script on path, exiting.')
script = result['message']
command = ['exec', script, '--layer', archive, '--extract', output_folder]
if not bot.is_quiet():
print("Extracting %s" % archive)
return run_command(command)
