def eks(ctx, profile):
check_kubernetes_client(ctx)
# Greet the user!
echo(
"Welcome to Metaflow! Follow the prompts to configure your "
"installation.\n",
bold=True,
)
check_kubernetes_config(ctx)
# Check for existing configuration.
if not overwrite_config(profile):
ctx.abort()
verify_aws_credentials(ctx)
existing_env = get_env(profile)
env = existing_env.copy()
if existing_env.get("METAFLOW_DEFAULT_DATASTORE") == "s3":
# Skip S3 configuration if it is already configured
pass
elif not existing_env.get("METAFLOW_DEFAULT_DATASTORE"):
env.update(configure_s3_datastore(existing_env))
else:
# If configured to use something else, offer to switch to S3
click.confirm(
"\nMetaflow on EKS needs to use S3 as a datastore, " +
"but your existing configuration is not using S3. " +
"Would you like to reconfigure it to use S3?",
default=True,
abort=True,
)
env.update(configure_s3_datastore(existing_env))
# Configure remote metadata.
if existing_env.get("METAFLOW_DEFAULT_METADATA") == "service":
# Skip metadata service configuration if it is already configured
pass
else:
if click.confirm(
"\nMetaflow can use a " +
yellow("remote Metadata Service to track") +
" and persist flow execution metadata. \nWould you like to "
"configure the Metadata Service?",
default=True,
abort=False,
):
env.update(configure_metadata_service(existing_env))
# Configure AWS EKS for compute.
env.update(configure_eks(existing_env))
persist_env({k: v for k, v in env.items() if v}, profile)
def pull(episode):
tutorials_dir = get_tutorials_dir()
if not episode:
episodes = get_all_episodes()
else:
episodes = [episode]
# Validate that the list is valid.
for episode in episodes:
validate_episode(episode)
# Create destination `metaflow-tutorials` dir.
dst_parent = os.path.join(os.getcwd(), "metaflow-tutorials")
makedirs(dst_parent)
# Pull specified episodes.
for episode in episodes:
dst_dir = os.path.join(dst_parent, episode)
# Check if episode has already been pulled before.
if os.path.exists(dst_dir):
if click.confirm("Episode " +
click.style('"{0}"'.format(episode), fg="red") +
" has already been pulled before. Do you wish "
"to delete the existing version?"):
shutil.rmtree(dst_dir)
else:
continue
echo("Pulling episode ", nl=False)
echo('"{0}"'.format(episode), fg="cyan", nl=False)
# TODO: Is the following redundant?
echo(" into your current working directory.")
# Copy from (local) metaflow package dir to current.
src_dir = os.path.join(tutorials_dir, episode)
shutil.copytree(src_dir, dst_dir)
echo("\nTo know more about an episode, type:\n", nl=False)
echo("metaflow tutorials info [EPISODE]", fg="cyan")
def reset(profile):
check_for_missing_profile(profile)
path = get_config_path(profile)
if os.path.exists(path):
if click.confirm(
"Do you really wish to reset the configuration in " +
click.style('"%s"' % path, fg="cyan"),
abort=True,
):
os.remove(path)
echo("Configuration successfully reset to run locally.")
else:
echo("Configuration is already reset to run locally.")
def check_kubernetes_config(ctx):
from kubernetes import config
try:
all_contexts, current_context = config.list_kube_config_contexts()
click.confirm(
"You have a valid kubernetes configuration. The current context is set to "
+ yellow(current_context["name"]) + " " + "Proceed?",
default=True,
abort=True,
)
except config.config_exception.ConfigException as e:
click.confirm(
"\nYou don't seem to have a valid kubernetes configuration file. "
+ "The error from kubernetes client library: " + red(str(e)) +
"." +
"To create a kubernetes configuration for EKS, you typically need to run "
+ yellow("aws eks update-kubeconfig --name <CLUSTER NAME>") +
". For further details, refer to AWS Documentation at https://docs.aws.amazon.com/eks/latest/userguide/create-kubeconfig.html\n"
"Do you want to proceed with configuring Metaflow for EKS anyway?",
default=False,
abort=True,
)
def configure_datastore_and_metadata(existing_env):
empty_profile = False
if not existing_env:
empty_profile = True
env = {}
# Configure Amazon S3 as the datastore.
use_s3_as_datastore = click.confirm(
"\nMetaflow can use " + yellow("Amazon S3 as the storage backend") +
" for all code and data artifacts on " +
"AWS.\nAmazon S3 is a strict requirement if you " +
"intend to execute your flows on AWS Batch " +
"and/or schedule them on AWS Step " +
"Functions.\nWould you like to configure Amazon " +
"S3 as the default storage backend?",
default=empty_profile
or existing_env.get("METAFLOW_DEFAULT_DATASTORE", "") == "s3",
abort=False,
)
if use_s3_as_datastore:
env.update(configure_s3_datastore(existing_env))
# Configure Metadata service for tracking.
if click.confirm(
"\nMetaflow can use a " +
yellow("remote Metadata Service to track") +
" and persist flow execution metadata.\nConfiguring the "
"service is a requirement if you intend to schedule your "
"flows with AWS Step Functions.\nWould you like to "
"configure the Metadata Service?",
default=empty_profile
or existing_env.get("METAFLOW_DEFAULT_METADATA", "") == "service"
or "METAFLOW_SFN_IAM_ROLE" in env,
abort=False,
):
env.update(configure_metadata_service(existing_env))
return env
def overwrite_config(profile):
path = get_config_path(profile)
if os.path.exists(path):
if not click.confirm(
click.style(
"We found an existing configuration for your " +
"profile. Do you want to modify the existing " +
"configuration?",
fg="red",
bold=True,
)):
echo(
"You can configure a different named profile by using the "
"--profile argument. You can activate this profile by setting "
"the environment variable METAFLOW_PROFILE to the named "
"profile.",
fg="yellow",
)
return False
return True
def verify_aws_credentials(ctx):
# Verify that the user has configured AWS credentials on their computer.
if not click.confirm(
"\nMetaflow relies on " + yellow("AWS access credentials") +
" present on your computer to access resources on AWS."
"\nBefore proceeding further, please confirm that you "
"have already configured these access credentials on "
"this computer.",
default=True,
):
echo(
"There are many ways to setup your AWS access credentials. You "
"can get started by following this guide: ",
nl=False,
fg="yellow",
)
echo(
"https://docs.aws.amazon.com/cli/latest/userguide/cli-chap-configure.html",
fg="cyan",
)
ctx.abort()
def aws(ctx, profile):
# Greet the user!
echo(
"Welcome to Metaflow! Follow the prompts to configure your "
"installation.\n",
bold=True,
)
# Check for existing configuration.
if not overwrite_config(profile):
ctx.abort()
verify_aws_credentials(ctx)
existing_env = get_env(profile)
empty_profile = False
if not existing_env:
empty_profile = True
env = {}
env.update(configure_datastore_and_metadata(existing_env))
# Configure AWS Batch for compute if using S3
if env.get("METAFLOW_DEFAULT_DATASTORE") == "s3":
if click.confirm(
"\nMetaflow can scale your flows by " +
yellow("executing your steps on AWS Batch") +
".\nAWS Batch is a strict requirement if you intend "
"to schedule your flows on AWS Step Functions.\nWould "
"you like to configure AWS Batch as your compute "
"backend?",
default=empty_profile
or "METAFLOW_BATCH_JOB_QUEUE" in existing_env,
abort=False,
):
env.update(configure_aws_batch(existing_env))
persist_env({k: v for k, v in env.items() if v}, profile)
def export(profile, output_filename):
check_for_missing_profile(profile)
# Export its contents to a new file.
path = get_config_path(profile)
env_dict = {}
if os.path.exists(path):
with open(path, "r") as f:
env_dict = json.load(f)
# resolve_path doesn't expand `~` in `path`.
output_path = expanduser(output_filename)
if os.path.exists(output_path):
if click.confirm(
"Do you wish to overwrite the contents in " +
click.style('"%s"' % output_path, fg="cyan") + "?",
abort=True,
):
pass
# Write to file.
with open(output_path, "w") as f:
json.dump(env_dict, f, indent=4, sort_keys=True)
echo("Configuration successfully exported to: ", nl=False)
echo('"%s"' % output_path, fg="cyan")
def configure_aws_batch(existing_env):
empty_profile = False
if not existing_env:
empty_profile = True
env = {}
# Set AWS Batch Job Queue.
env["METAFLOW_BATCH_JOB_QUEUE"] = click.prompt(
cyan("[METAFLOW_BATCH_JOB_QUEUE]") + " AWS Batch Job Queue.",
default=existing_env.get("METAFLOW_BATCH_JOB_QUEUE"),
show_default=True,
)
# Set IAM role for AWS Batch jobs to assume.
env["METAFLOW_ECS_S3_ACCESS_IAM_ROLE"] = click.prompt(
cyan("[METAFLOW_ECS_S3_ACCESS_IAM_ROLE]") +
" IAM role for AWS Batch jobs to access AWS " +
"resources (Amazon S3 etc.).",
default=existing_env.get("METAFLOW_ECS_S3_ACCESS_IAM_ROLE"),
show_default=True,
)
# Set default Docker repository for AWS Batch jobs.
env["METAFLOW_BATCH_CONTAINER_REGISTRY"] = click.prompt(
cyan("[METAFLOW_BATCH_CONTAINER_REGISTRY]") + yellow(" (optional)") +
" Default Docker image repository for AWS " +
"Batch jobs. If nothing is specified, " +
"dockerhub (hub.docker.com/) is " + "used as default.",
default=existing_env.get("METAFLOW_BATCH_CONTAINER_REGISTRY", ""),
show_default=True,
)
# Set default Docker image for AWS Batch jobs.
env["METAFLOW_BATCH_CONTAINER_IMAGE"] = click.prompt(
cyan("[METAFLOW_BATCH_CONTAINER_IMAGE]") + yellow(" (optional)") +
" Default Docker image for AWS Batch jobs. " +
"If nothing is specified, an appropriate " +
"python image is used as default.",
default=existing_env.get("METAFLOW_BATCH_CONTAINER_IMAGE", ""),
show_default=True,
)
# Configure AWS Step Functions for scheduling.
if click.confirm(
"\nMetaflow can " + yellow("schedule your flows on AWS Step "
"Functions") +
" and trigger them at a specific cadence using "
"Amazon EventBridge.\nTo support flows involving "
"foreach steps, you would need access to AWS "
"DynamoDB.\nWould you like to configure AWS Step "
"Functions for scheduling?",
default=empty_profile or "METAFLOW_SFN_IAM_ROLE" in existing_env,
abort=False,
):
# Configure IAM role for AWS Step Functions.
env["METAFLOW_SFN_IAM_ROLE"] = click.prompt(
cyan("[METAFLOW_SFN_IAM_ROLE]") +
" IAM role for AWS Step Functions to " +
"access AWS resources (AWS Batch, " + "AWS DynamoDB).",
default=existing_env.get("METAFLOW_SFN_IAM_ROLE"),
show_default=True,
)
# Configure IAM role for AWS Events Bridge.
env["METAFLOW_EVENTS_SFN_ACCESS_IAM_ROLE"] = click.prompt(
cyan("[METAFLOW_EVENTS_SFN_ACCESS_IAM_ROLE]") +
" IAM role for Amazon EventBridge to " +
"access AWS Step Functions.",
default=existing_env.get("METAFLOW_EVENTS_SFN_ACCESS_IAM_ROLE"),
show_default=True,
)
# Configure AWS DynamoDB Table for AWS Step Functions.
env["METAFLOW_SFN_DYNAMO_DB_TABLE"] = click.prompt(
cyan("[METAFLOW_SFN_DYNAMO_DB_TABLE]") +
" AWS DynamoDB table name for tracking " +
"AWS Step Functions execution metadata.",
default=existing_env.get("METAFLOW_SFN_DYNAMO_DB_TABLE"),
show_default=True,
)
return env
