def get_manifest_data(bucket, team, dataset, manifest_key):
""" Returns a list of items from manifests control table """
dynamo_config = DynamoConfiguration()
dynamo_interface = DynamoInterface(dynamo_config)
s3_interface = S3Interface()
local_path = s3_interface.download_object(bucket, manifest_key)
ddb_keys = []
items = []
with open(local_path, "r") as raw_file:
file_names = [
file_name.strip().split("/")[-1] for file_name in raw_file
]
for file in file_names:
ddb_keys.append({
"dataset_name": team + "-" + dataset,
"manifest_file_name": manifest_key.split("/")[-1],
"datafile_name": file
})
for ddb_key in ddb_keys:
try:
items.append(
dynamo_interface.get_item_from_manifests_control_table(
ddb_key["dataset_name"], ddb_key["manifest_file_name"],
ddb_key["datafile_name"]))
except KeyError:
logger.error("The manifest file has not been processed in Stage A")
raise Exception("Manifest File has not been processed in Stage A")
return items
def get_ddb_keys(keys_to_process, bucket, team, dataset):
### Returns a list of DynamoDB keys for Querying
ddb_keys = []
for key in keys_to_process:
s3_interface = S3Interface()
local_path = s3_interface.download_object(bucket, key)
with open(local_path, "r") as raw_file:
file_names = [
file_name.strip().split("/")[-1] for file_name in raw_file
]
for file in file_names:
ddb_keys.append({
"dataset_name": team + "-" + dataset,
"manifest_file_name": key.split("/")[-1],
"datafile_name": file
})
return ddb_keys
def lambda_handler(event, context):
"""Updates the S3 objects metadata catalog
Arguments:
event {dict} -- Dictionary with details on Bucket and Keys
context {dict} -- Dictionary with details on Lambda context
Returns:
{dict} -- Dictionary with response
"""
try:
logger.info('Fetching event data from previous step')
bucket = event['body']['bucket']
processed_keys_path = event['body']['job']['processedKeysPath']
processed_keys = S3Interface().list_objects(bucket,
processed_keys_path)
team = event['body']['team']
pipeline = event['body']['pipeline']
stage = event['body']['pipeline_stage']
dataset = event['body']['dataset']
peh_id = event['body']['job']['peh_id']
keys_to_process = event['body']['keysToProcess']
s3_path = "post-stage/{}/manifests/{}/{}".format(
team, dataset, keys_to_process[0].split("/")[-1])
logger.info('Initializing Octagon client')
component = context.function_name.split('-')[-2].title()
octagon_client = (octagon.OctagonClient().with_run_lambda(
True).with_configuration_instance(event['body']['env']).build())
peh.PipelineExecutionHistoryAPI(
octagon_client).retrieve_pipeline_execution(peh_id)
logger.info('Initializing DynamoDB config and Interface')
dynamo_config = DynamoConfiguration()
dynamo_interface = DynamoInterface(dynamo_config)
logger.info('Storing metadata to DynamoDB')
for key in processed_keys:
object_metadata = {
'bucket': bucket,
'key': key,
'size': S3Interface().get_size(bucket, key),
'last_modified_date':
S3Interface().get_last_modified(bucket, key),
'org': event['body']['org'],
'app': event['body']['app'],
'env': event['body']['env'],
'team': team,
'pipeline': pipeline,
'dataset': dataset,
'stage': 'stage',
'pipeline_stage': stage,
'peh_id': peh_id
}
dynamo_interface.update_object_metadata_catalog(object_metadata)
logger.info("Updating manifests control table")
items = get_manifest_data(bucket, team, dataset, keys_to_process[0])
ddb_keys = get_ddb_keys(items)
for ddb_key in ddb_keys:
dynamo_interface.update_manifests_control_table_stageb(
ddb_key, "COMPLETED")
logger.info("Move manifest file to post stage")
kms_key = KMSConfiguration(team).get_kms_arn
s3_interface = S3Interface()
s3_interface.copy_object(bucket,
keys_to_process[0],
bucket,
s3_path,
kms_key=kms_key)
logger.info("Removing manifest file from pre-stage")
s3_interface.delete_objects(bucket, keys_to_process[0])
# Only uncomment if a queue for the next stage exists
# logger.info('Sending messages to next SQS queue if it exists')
# sqs_config = SQSConfiguration(team, dataset, ''.join([stage[:-1], chr(ord(stage[-1]) + 1)]))
# sqs_interface = SQSInterface(sqs_config.get_stage_queue_name)
# sqs_interface.send_batch_messages_to_fifo_queue(processed_keys, 10, '{}-{}'.format(team, dataset))
octagon_client.update_pipeline_execution(
status="{} {} Processing".format(stage, component),
component=component)
octagon_client.end_pipeline_execution_success()
except Exception as e:
logger.error("Fatal error", exc_info=True)
octagon_client.end_pipeline_execution_failed(
component=component,
issue_comment="{} {} Error: {}".format(stage, component, repr(e)))
for ddb_key in ddb_keys:
dynamo_interface.update_manifests_control_table_stageb(
ddb_key, "FAILED", None, "Failed in Post Update")
raise e
return 200
import boto3
import time
import sys
import logging
import traceback
import string
import random
#######################################################
# Use S3 Interface to interact with S3 objects
# For example to download/upload them
#######################################################
from datalake_library.commons import init_logger
from datalake_library.configuration.resource_configs import S3Configuration, KMSConfiguration
from datalake_library.interfaces.s3_interface import S3Interface
s3_interface = S3Interface()
# IMPORTANT: Stage bucket where transformed data must be uploaded
stage_bucket = S3Configuration().stage_bucket
athena_client = boto3.client('athena')
glue_client = boto3.client('glue')
logger = logging.getLogger()
logger.setLevel(logging.DEBUG)
class CustomTransform():
def __init__(self):
logger.info("Athena Light Transform initiated")
def transform_object(self, bucket, body, team, dataset):
# returns table path, or table path with partition name
def lambda_handler(event, context):
"""Updates the S3 objects metadata catalog
Arguments:
event {dict} -- Dictionary with details on previous processing step
context {dict} -- Dictionary with details on Lambda context
Returns:
{dict} -- Dictionary with outcome of the process
"""
try:
logger.info('Fetching event data from previous step')
processed_keys = event['body']['processedKeys']
team = event['body']['team']
pipeline = event['body']['pipeline']
stage = event['body']['pipeline_stage']
dataset = event['body']['dataset']
peh_id = event['body']['peh_id']
logger.info('Initializing Octagon client')
component = context.function_name.split('-')[-2].title()
octagon_client = (octagon.OctagonClient().with_run_lambda(
True).with_configuration_instance(event['body']['env']).build())
peh.PipelineExecutionHistoryAPI(
octagon_client).retrieve_pipeline_execution(peh_id)
logger.info('Initializing DynamoDB config and Interface')
dynamo_config = DynamoConfiguration()
dynamo_interface = DynamoInterface(dynamo_config)
logger.info('Storing metadata to DynamoDB')
bucket = S3Configuration().stage_bucket
for key in processed_keys:
object_metadata = {
'bucket': bucket,
'key': key,
'size': S3Interface().get_size(bucket, key),
'last_modified_date':
S3Interface().get_last_modified(bucket, key),
'org': event['body']['org'],
'app': event['body']['app'],
'env': event['body']['env'],
'team': team,
'pipeline': pipeline,
'dataset': dataset,
'stage': 'stage',
'pipeline_stage': stage,
'peh_id': peh_id
}
dynamo_interface.update_object_metadata_catalog(object_metadata)
#Workload management changes
#---------------------------
wlm_ddb_table = dynamo_interface.wlm_control_table
item = dynamo_interface.get_item(
wlm_ddb_table, {
"name":
"{}-{}-{}".format(team, dataset,
processed_keys[0].split("/")[-2])
})
priority = item.get('priority', None)
print(priority)
#---------------------------
logger.info('Sending messages to next SQS queue if it exists')
sqs_config = SQSConfiguration(team, dataset, ''.join(
[stage[:-1], chr(ord(stage[-1]) + 1)]),
priority) #Workload management changes
sqs_interface = SQSInterface(
sqs_config.get_stage_queue_name_wlm) #Workload management changes
sqs_interface.send_batch_messages_to_fifo_queue(
processed_keys, 10, '{}-{}'.format(team, dataset))
octagon_client.update_pipeline_execution(
status="{} {} Processing".format(stage, component),
component=component)
octagon_client.end_pipeline_execution_success()
except Exception as e:
logger.error("Fatal error", exc_info=True)
octagon_client.end_pipeline_execution_failed(
component=component,
issue_comment="{} {} Error: {}".format(stage, component, repr(e)))
raise e
return 200
def lambda_handler(event, context):
"""Updates the S3 objects metadata catalog
Arguments:
event {dict} -- Dictionary with details on Bucket and Keys
context {dict} -- Dictionary with details on Lambda context
Returns:
{dict} -- Dictionary with response
"""
def replace_decimals(obj):
if isinstance(obj, list):
for i in range(len(obj)):
obj[i] = replace_decimals(obj[i])
return obj
elif isinstance(obj, dict):
for k, v in obj.items():
obj[k] = replace_decimals(v)
return obj
elif isinstance(obj, set):
return set(replace_decimals(i) for i in obj)
elif isinstance(obj, decimal.Decimal):
if obj % 1 == 0:
return int(obj)
else:
return float(obj)
else:
return obj
def get_table_partitions(db, tbl):
glue_response = glue_client.get_table(DatabaseName=db, Name=tbl)
logger.debug('Glue get_table response: {}'.format(glue_response))
return glue_response['Table']['PartitionKeys']
try:
logger.info('Fetching event data from previous step')
bucket = event['body']['bucket']
processed_keys_path = event['body']['job']['processedKeysPath']
processed_keys = S3Interface().list_objects(bucket,
processed_keys_path)
team = event['body']['team']
pipeline = event['body']['pipeline']
stage = event['body']['pipeline_stage']
dataset1 = event['body']['dataset']
peh_id = event['body']['job']['peh_id']
env = event['body']['env']
logger.info('Initializing Octagon client')
component = context.function_name.split('-')[-2].title()
octagon_client = (octagon.OctagonClient().with_run_lambda(
True).with_configuration_instance(event['body']['env']).build())
peh.PipelineExecutionHistoryAPI(
octagon_client).retrieve_pipeline_execution(peh_id)
logger.info('Initializing DynamoDB config and Interface')
dynamo_config = DynamoConfiguration()
dynamo_interface = DynamoInterface(dynamo_config)
logger.info(
'Storing metadata to DynamoDB and tagging resulting S3 Objects')
for key in processed_keys:
object_metadata = {
'bucket': bucket,
'key': key,
'size': S3Interface().get_size(bucket, key),
'last_modified_date':
S3Interface().get_last_modified(bucket, key),
'org': event['body']['org'],
'app': event['body']['app'],
'env': event['body']['env'],
'team': team,
'pipeline': pipeline,
'dataset': dataset1,
'stage': 'stage',
'pipeline_stage': stage,
'peh_id': peh_id
}
dynamo_interface.update_object_metadata_catalog(object_metadata)
tag_keys = ['org', 'app', 'env', 'team', 'dataset']
tag_dict = {key: object_metadata[key] for key in tag_keys}
S3Interface().tag_object(bucket, key, tag_dict)
# Only uncomment if a queue for the next stage exists
# logger.info('Sending messages to next SQS queue if it exists')
# sqs_config = SQSConfiguration(team, dataset, ''.join([stage[:-1], chr(ord(stage[-1]) + 1)]))
# sqs_interface = SQSInterface(sqs_config.get_stage_queue_name)
# sqs_interface.send_batch_messages_to_fifo_queue(processed_keys, 10, '{}-{}'.format(team, dataset))
prestage_table = event['body']['dest_table']['name']
prestage_db = event['body']['dest_db']
dest_part_name = event['body']['dest_table']['part_name']
dest_part_value = event['body']['dest_table']['part_value']
processOutput = {}
if dest_part_name is not '' and dest_part_value is not '':
partitions = []
part_dict = {"name": dest_part_name, "value": dest_part_value}
partitions.append(part_dict)
processOutput['partitions'] = partitions
processOutput['processed_keys'] = processed_keys
ssmresponse = ssmcli.get_parameter(
Name=f'/SDLF/DDB/{team}/{pipeline}/DependenciesByTable')
ddb_dependencies_by_table = ssmresponse['Parameter']['Value']
ddb_table = dynamodb.Table(ddb_dependencies_by_table)
ssmresponse = ssmcli.get_parameter(
Name=f'/SDLF/DDB/{team}/{pipeline}/Dependencies')
ddb_dependencies = ssmresponse['Parameter']['Value']
consulta = f'{prestage_db.lower()}.{prestage_table.lower()}'
logger.info(consulta)
response = ddb_table.get_item(Key={'table_name': consulta})
logger.info(f'Response {response}')
if 'Item' in response:
list_transforms = response['Item']['list_transforms']
num_of_transforms = len(list_transforms)
logger.debug(f'Response {response}')
logger.info(f'This table triggers {num_of_transforms} datasets')
next_stage = 'B'
stage_b_message = {}
for dataset in list_transforms:
ddb_steps = dynamodb.Table(ddb_dependencies)
logger.info(dataset)
response = ddb_steps.get_item(Key={'dataset': dataset})
logger.info(f'Response {response}')
num_of_transforms = len(list_transforms)
item = response['Item']
dest_table = item['dataset'].split('.')[1]
dest_db = item['dataset'].split('.')[0]
dependencies = item['dependencies']
date_substitutions = replace_decimals(
item.get('date_substitutions', []))
logger.info(f'Dependencies: {dependencies}')
partition = item.get('partitionColumn', '')
partition_mask = item.get('partitionPythonMask', None)
partition_value_formatted = None
table_check = []
for table in dependencies:
table_name = table['TableName'].split('.')[1]
table_db = table['TableName'].split('.')[0]
table_partition = table.get('FieldColumn', '')
table_partition_format = table.get('DateExpression', None)
relativedelta_attributes = replace_decimals(
table.get('relativedelta_attributes', None))
table_partitions = processOutput.get('partitions', [])
usage = table.get('Usage', 'validate').lower()
if usage == 'validate':
if prestage_db == table_db and prestage_table == table_name:
logger.info(
f'This table does not update/overwrite {dataset} dataset'
)
break
else:
logger.debug(
f'Table {table_db}.{table_name} is not the trigger table'
)
else:
if prestage_db.lower() == table_db.lower(
) and prestage_table.lower() == table_name.lower():
# dst_tbl_partitions = get_table_partitions(prestage_db,prestage_table)
partition_value_formatted = ''
# If dest table has partitions and source table has partitions
logger.debug(
f'Partition: {partition}, table_partitions: {table_partitions}'
)
if table_partitions and table_partition_format is not None:
table_partition_value = table_partitions[0][
'value']
value = datetime.strptime(
table_partition_value,
table_partition_format)
target_value = value - relativedelta(
**relativedelta_attributes)
partition_value_formatted = target_value.strftime(
partition_mask)
logger.info(
f'This table {usage.upper()} dataset {dest_table} '
f' Partition {partition} = {partition_value_formatted}'
)
# validate(table_db, table_name, table_partitions)
stage_b_message[
'prev_stage_processed_keys'] = processed_keys
stage_b_message['team'] = team
stage_b_message['pipeline'] = pipeline
stage_b_message['pipeline_stage'] = ''.join(
[stage[:-1], next_stage])
stage_b_message['dataset'] = dataset1
stage_b_message['org'] = event['body']['org']
stage_b_message['app'] = event['body']['app']
stage_b_message['env'] = event['body']['env']
stage_b_message['behaviour'] = table[
'Usage'].lower()
stage_b_message['dest_db'] = dest_db
stage_b_message['dest_table'] = {}
stage_b_message['dest_table']['name'] = dest_table
stage_b_message['dest_table'][
'part_name'] = partition
stage_b_message['dest_table'][
'part_value'] = partition_value_formatted
stage_b_message['steps'] = item['steps']
stage_b_message[
'date_substitutions'] = date_substitutions
logger.info(
'Sending messages to next SQS queue if it exists'
)
# GEt queue by SSM
logger.info(stage_b_message)
sqs_config = SQSConfiguration(
team, pipeline, stage)
sqs_interface = SQSInterface(
sqs_config.get_stage_queue_name)
sqs_interface.send_message_to_fifo_queue(
json.dumps(stage_b_message),
'{}-{}'.format(team, pipeline))
break
else:
logger.info(f'This table triggers 0 datasets')
octagon_client.update_pipeline_execution(
status=f'{stage} {component} Processing', component=component)
octagon_client.end_pipeline_execution_success()
except Exception as e:
logger.error("Fatal error", exc_info=True)
octagon_client.end_pipeline_execution_failed(
component=component,
issue_comment=f'{stage} {component} Error: {repr(e)}')
raise e
return 200
def lambda_handler(event, context):
""" Load Datafile metadata in manifests control table
Check if manifest file is available within the threshold
Arguments:
event {dict} -- Dictionary with details on previous processing step
context {dict} -- Dictionary with details on Lambda context
Returns:
{dict} -- Dictionary with outcome of the process
"""
s3_interface = S3Interface()
stage_bucket = S3Configuration().stage_bucket
dynamo_config = DynamoConfiguration()
dynamo_interface = DynamoInterface(dynamo_config)
current_time = dt.datetime.utcnow()
current_timestamp = current_time.timestamp()
try:
logger.info("Fetching event data from previous step")
team = event['body']['team']
pipeline = event['body']['pipeline']
stage = event['body']['pipeline_stage']
dataset = event['body']['dataset']
peh_id = event['body']['peh_id']
env = event['body']['env']
bucket = event['body']['bucket']
input_file_key = event['body']['key']
input_file_name = input_file_key.split("/")[-1]
manifest_file_pattern = event['body']['manifest_details'][
'regex_pattern']
manifest_timeout = int(
event['body']['manifest_details']['manifest_timeout'])
if 'manifest_interval' in event['body']:
manifest_interval = event['body']['manifest_interval']
else:
manifest_interval = current_timestamp
logger.info('Initializing Octagon client')
component = context.function_name.split('-')[-2].title()
octagon_client = (octagon.OctagonClient().with_run_lambda(
True).with_configuration_instance(env).build())
peh.PipelineExecutionHistoryAPI(
octagon_client).retrieve_pipeline_execution(peh_id)
octagon_client.update_pipeline_execution(
status="{} {} Processing".format(stage, component),
component=component)
### List S3 Objects for the manifest file in the manifest prefix
### For this to work the manifest should have been loaded into DynamoDB
manifest_key = "pre-stage/{}/manifests/{}/".format(team, dataset)
processed_manifest_keys = s3_interface.list_objects(
stage_bucket, manifest_key)
matched_keys = []
items = []
if not processed_manifest_keys:
logger.info(
"Manifest File has not been loaded, sleeping for 5 mins")
time.sleep(300)
manifest_file_loaded = "False"
else:
for manifest_file_key in processed_manifest_keys:
manifest_file_name = manifest_file_key.split("/")[-1]
match = re.match(manifest_file_pattern, manifest_file_name)
if match:
matched_keys.append(manifest_file_name)
### Query Manifests Control table
for keys in matched_keys:
dataset_name = team + "-" + dataset
try:
items.append(
dynamo_interface.
get_item_from_manifests_control_table(
dataset_name, keys, input_file_name))
except KeyError:
logger.info(
"Manifest File has not been loaded, sleeping for 5 mins"
)
manifest_file_loaded = "False"
### Update Manifests Control table
if not items:
logger.info(
"Manifest File has not been loaded, sleeping for 5 mins"
)
time.sleep(300)
manifest_file_loaded = "False"
else:
ddb_key = {
'dataset_name': items[0]['dataset_name'],
'datafile_name': items[0]['datafile_name']
}
STATUS = "STARTED"
dynamo_interface.update_manifests_control_table_stagea(
ddb_key, STATUS)
manifest_file_loaded = "True"
event['body']['manifest_ddb_key'] = ddb_key
### Check if Manifest threshold has exceeded
if current_timestamp == manifest_interval:
current_timestamp = dt.datetime.utcnow().timestamp()
if int(
(current_timestamp - manifest_interval) / 60) >= manifest_timeout:
logger.error("Manifest Threshold Breached")
raise Exception("Manifest Threshold Breached")
event['body']['manifest_interval'] = manifest_interval
event['body']['manifest_file_loaded'] = manifest_file_loaded
except Exception as e:
logger.error("Fatal error", exc_info=True)
octagon_client.end_pipeline_execution_failed(
component=component,
issue_comment="{} {} Error: {}".format(stage, component, repr(e)))
raise e
return event
def lambda_handler(event, context):
""" Process the manifest file and loads into DynamoDB
Arguments:
event {dict} -- Dictionary with details on previous processing step
context {dict} -- Dictionary with details on Lambda context
Returns:
{dict} -- Dictionary with outcome of the process
"""
s3_interface = S3Interface()
stage_bucket = S3Configuration().stage_bucket
dynamo_config = DynamoConfiguration()
dynamo_interface = DynamoInterface(dynamo_config)
try:
logger.info("Fetching event data from previous step")
team = event['body']['team']
pipeline = event['body']['pipeline']
stage = event['body']['pipeline_stage']
dataset = event['body']['dataset']
peh_id = event['body']['peh_id']
env = event['body']['env']
bucket = event['body']['bucket']
manifest_file_key = event['body']['key']
manifest_file_name = manifest_file_key.split("/")[-1]
logger.info('Initializing Octagon client')
component = context.function_name.split('-')[-2].title()
octagon_client = (octagon.OctagonClient().with_run_lambda(
True).with_configuration_instance(env).build())
peh.PipelineExecutionHistoryAPI(
octagon_client).retrieve_pipeline_execution(peh_id)
### Download the manifest file to local
local_path = s3_interface.download_object(bucket, manifest_file_key)
### Process the manifest file
with open(local_path, "r") as raw_file:
file_names = [
file_name.strip().split("/")[-1] for file_name in raw_file
]
### Load data into manifests control table
for file in file_names:
item = {
"dataset_name":
team + "-" + dataset + "-" + manifest_file_name,
"datafile_name": manifest_file_name + "-" + file
}
dynamo_interface.put_item_in_manifests_control_table(item)
### Set s3 path for Copy
s3_path = 'pre-stage/{}/manifests/{}/{}'.format(
team, dataset, manifest_file_name)
kms_key = KMSConfiguration(team).get_kms_arn
### Copy Manifest File to team/manifest/dataset location
s3_interface.copy_object(bucket,
manifest_file_key,
stage_bucket,
s3_path,
kms_key=kms_key)
octagon_client.update_pipeline_execution(
status="{} {} Processing".format(stage, component),
component=component)
processed_keys = [s3_path]
except Exception as e:
logger.error("Fatal error", exc_info=True)
octagon_client.end_pipeline_execution_failed(
component=component,
issue_comment="{} {} Error: {}".format(stage, component, repr(e)))
raise e
return processed_keys
def lambda_handler(event, context):
"""Updates the S3 objects metadata catalog
Arguments:
event {dict} -- Dictionary with details on Bucket and Keys
context {dict} -- Dictionary with details on Lambda context
Returns:
{dict} -- Dictionary with response
"""
try:
logger.info('Fetching event data from previous step')
bucket = event['body']['bucket']
processed_keys_path = event['body']['job']['processedKeysPath']
processed_keys = S3Interface().list_objects(bucket,
processed_keys_path)
team = event['body']['team']
pipeline = event['body']['pipeline']
stage = event['body']['pipeline_stage']
dataset = event['body']['dataset']
logger.info('Initializing Octagon client')
component = context.function_name.split('-')[-2].title()
octagon_client = (octagon.OctagonClient().with_run_lambda(
True).with_configuration_instance(event['body']['env']).build())
peh.PipelineExecutionHistoryAPI(
octagon_client).retrieve_pipeline_execution(
event['body']['job']['peh_id'])
logger.info('Initializing DynamoDB config and Interface')
dynamo_config = DynamoConfiguration()
dynamo_interface = DynamoInterface(dynamo_config)
logger.info('Storing metadata to DynamoDB')
for key in processed_keys:
object_metadata = {
'bucket': bucket,
'key': key,
'team': team,
'pipeline': pipeline,
'dataset': dataset,
'peh_id': event['body']['job']['peh_id'],
'stage': 'post-stage'
}
dynamo_interface.update_object_metadata_catalog(object_metadata)
# Add Tables to Result Path to Enable Deequ Job
table_path = "compile_topics_data_csv"
tables = [table_path]
# Only uncomment if using Kendra and index and data source ALREADY created
# Data Sync Job
# kendra_client = boto3.client('kendra')
# response = kendra_client.start_data_source_sync_job(
# Id='ENTER_DATASOURCE_ID',
# IndexId='ENTER_INDEX_ID''
# )
# Only uncomment if a queue for the next stage exists
# logger.info('Sending messages to next SQS queue if it exists')
# sqs_config = SQSConfiguration(team, dataset, ''.join([stage[:-1], chr(ord(stage[-1]) + 1)]))
# sqs_interface = SQSInterface(sqs_config.get_stage_queue_name)
# sqs_interface.send_batch_messages_to_fifo_queue(processed_keys, 10, '{}-{}'.format(team, dataset))
octagon_client.update_pipeline_execution(
status="{} {} Processing".format(stage, component),
component=component)
octagon_client.end_pipeline_execution_success()
except Exception as e:
logger.error("Fatal error", exc_info=True)
octagon_client.end_pipeline_execution_failed(
component=component,
issue_comment="{} {} Error: {}".format(stage, component, repr(e)))
raise e
return tables
def lambda_handler(event, context):
"""Updates the S3 objects metadata catalog
Arguments:
event {dict} -- Dictionary with details on Bucket and Keys
context {dict} -- Dictionary with details on Lambda context
Returns:
{dict} -- Dictionary with response
"""
try:
logger.info('Fetching event data from previous step')
bucket = event['body']['bucket']
processed_keys_path = event['body']['job']['processedKeysPath']
processed_keys = S3Interface().list_objects(bucket,
processed_keys_path)
team = event['body']['team']
pipeline = event['body']['pipeline']
stage = event['body']['pipeline_stage']
dataset = event['body']['dataset']
# Lastly, Lets Run Comprehend Multi-Label Classification Job
# with the Training Data we created earlier:
# Connect to Comprehend Client
comprehend_client = boto3.client('comprehend')
# Set Parameters for Classifier Training Job and get KMS Key for Encryption
input_key = "post-stage/{}/{}/multilabel_classification/training_data.csv".format(
team, dataset)
s3_input = "s3://{}/{}".format(bucket, input_key)
output_key = "post-stage/{}/{}/multilabel_classification/".format(
team, dataset)
s3_output = "s3://{}/{}".format(bucket, output_key)
kms_key = KMSConfiguration(team).get_kms_arn
name = "MedicalResearchTopicClassifier"
aws_account_id = context.invoked_function_arn.split(":")[4]
data_access_role = 'arn:aws:iam::{}:role/sdlf-{}-{}-create-classifier-b'.format(
aws_account_id, team, pipeline)
# Call Multi-Label Classifier Training to Start
response = comprehend_client.create_document_classifier(
DocumentClassifierName=name,
DataAccessRoleArn=data_access_role,
Tags=[
{
'Key': 'Framework',
'Value': 'sdlf'
},
],
InputDataConfig={
'S3Uri': s3_input,
'LabelDelimiter': '|'
},
OutputDataConfig={
'S3Uri': s3_output,
'KmsKeyId': kms_key
},
LanguageCode='en',
VolumeKmsKeyId=kms_key,
Mode='MULTI_LABEL')
except Exception as e:
logger.error("Fatal error", exc_info=True)
raise e
return 200