问题
I upload 10 files every day at 11 p.m with a Cron Job to a bucket on GCS. Each file is a .csv with a size from 2 to 30 KB. The file name is always YYYY-MM-DD-ID.csv
A Cloud Function is called everytime I am uploading a file into that bucket to send those .csv files to BigQuery. The trigger type is Cloud Storage on finalise/create events.
My issue is the following: On BigQuery, each value for each row/columns is multiplied by a multiple. Sometimes it's 1 (so the value is the same), often 2 and sometimes 3. I attached one example bellow with the difference between BigQuery (BQ) and Google Cloud Storage (GCS).
It seems that the cloud function is called multiple times. It's not on the code but rather some duplicate message deliveries from the Cloud Function during the trigger. When I am going o the logs tab for today, I can see the Cloud Function upload_to_bigquery has been called multiple times.
I have tried to fix it but I made a mistake. I thought we could write temporary files to Cloud Functions but we can not. My solution was to write the filename I am uploading to BigQuery on a .txt file. And before to upload a new file on BigQuery, read that .txt file and check if the current file exist on that list. If the filename is already present, skip. Else, write the .txt filename to the list and do my stuff.
if file_to_upload not in text:
text.append(file_to_upload)
with open("all_uploaded_files.txt", "w") as text_file:
for item in text:
text_file.write(item + "\n")
bucket = storage_client.bucket('sfr-test-data')
blob = bucket.blob("all_uploaded_files.txt")
blob.upload_from_filename("all_uploaded_files.txt")
## do my things here
else:
print("file already uploaded")
# skip to new file to upload
But even if I could do that, this solution is not viable. The temporary file will become so large after months of years that it would be a mess. Do you know whats the easiest way to fix this issue?
Cloud Function: upload_to_big_query - main.py
BUCKET = "xxx"
GOOGLE_PROJECT = "xxx"
HEADER_MAPPING = {
"Source/Medium": "source_medium",
"Campaign": "campaign",
"Last Non-Direct Click Conversions": "last_non_direct_click_conversions",
"Last Non-Direct Click Conversion Value": "last_non_direct_click_conversion_value",
"Last Click Prio Conversions": "last_click_prio_conversions",
"Last Click Prio Conversion Value": "last_click_prio_conversion_value",
"Data-Driven Conversions": "dda_conversions",
"Data-Driven Conversion Value": "dda_conversion_value",
"% Change in Conversions from Last Non-Direct Click to Last Click Prio": "last_click_prio_vs_last_click",
"% Change in Conversions from Last Non-Direct Click to Data-Driven": "dda_vs_last_click"
}
SPEND_HEADER_MAPPING = {
"Source/Medium": "source_medium",
"Campaign": "campaign",
"Spend": "spend"
}
tables_schema = {
"google-analytics": [
bigquery.SchemaField("date", bigquery.enums.SqlTypeNames.DATE, mode='REQUIRED'),
bigquery.SchemaField("week", bigquery.enums.SqlTypeNames.INT64, mode='REQUIRED'),
bigquery.SchemaField("goal", bigquery.enums.SqlTypeNames.STRING, mode='REQUIRED'),
bigquery.SchemaField("source", bigquery.enums.SqlTypeNames.STRING, mode='NULLABLE'),
bigquery.SchemaField("medium", bigquery.enums.SqlTypeNames.STRING, mode='NULLABLE'),
bigquery.SchemaField("campaign", bigquery.enums.SqlTypeNames.STRING, mode='NULLABLE'),
bigquery.SchemaField("last_non_direct_click_conversions", bigquery.enums.SqlTypeNames.INT64, mode='NULLABLE'),
bigquery.SchemaField("last_non_direct_click_conversion_value", bigquery.enums.SqlTypeNames.FLOAT64, mode='NULLABLE'),
bigquery.SchemaField("last_click_prio_conversions", bigquery.enums.SqlTypeNames.INT64, mode='NULLABLE'),
bigquery.SchemaField("last_click_prio_conversion_value", bigquery.enums.SqlTypeNames.FLOAT64, mode='NULLABLE'),
bigquery.SchemaField("dda_conversions", bigquery.enums.SqlTypeNames.FLOAT64, mode='NULLABLE'),
bigquery.SchemaField("dda_conversion_value", bigquery.enums.SqlTypeNames.FLOAT64, mode='NULLABLE'),
bigquery.SchemaField("last_click_prio_vs_last_click", bigquery.enums.SqlTypeNames.FLOAT64, mode='NULLABLE'),
bigquery.SchemaField("dda_vs_last_click", bigquery.enums.SqlTypeNames.FLOAT64, mode='NULLABLE')
],
"google-analytics-spend": [
bigquery.SchemaField("date", bigquery.enums.SqlTypeNames.DATE, mode='REQUIRED'),
bigquery.SchemaField("week", bigquery.enums.SqlTypeNames.INT64, mode='REQUIRED'),
bigquery.SchemaField("source", bigquery.enums.SqlTypeNames.STRING, mode='NULLABLE'),
bigquery.SchemaField("medium", bigquery.enums.SqlTypeNames.STRING, mode='NULLABLE'),
bigquery.SchemaField("campaign", bigquery.enums.SqlTypeNames.STRING, mode='NULLABLE'),
bigquery.SchemaField("spend", bigquery.enums.SqlTypeNames.FLOAT64, mode='NULLABLE'),
]
}
def download_from_gcs(file):
client = storage.Client()
bucket = client.get_bucket(BUCKET)
blob = bucket.get_blob(file['name'])
file_name = os.path.basename(os.path.normpath(file['name']))
blob.download_to_filename(f"/tmp/{file_name}")
return file_name
def load_in_bigquery(file_object, dataset: str, table: str):
client = bigquery.Client()
table_id = f"{GOOGLE_PROJECT}.{dataset}.{table}"
job_config = bigquery.LoadJobConfig(
source_format=bigquery.SourceFormat.CSV,
skip_leading_rows=1,
autodetect=True,
schema=tables_schema[table]
)
job = client.load_table_from_file(file_object, table_id, job_config=job_config)
job.result() # Wait for the job to complete.
def __order_columns(df: pd.DataFrame, spend=False) ->pd.DataFrame:
# We want to have source and medium columns at the third position
# for a spend data frame and at the fourth postion for others df
# because spend data frame don't have goal column.
pos = 2 if spend else 3
cols = df.columns.tolist()
cols[pos:2] = cols[-2:]
cols = cols[:-2]
return df[cols]
def __common_transformation(df: pd.DataFrame, date: str, goal: str) -> pd.DataFrame:
# for any kind of dataframe, we add date and week columns
# based on the file name and we split Source/Medium from the csv
# into two different columns
week_of_the_year = datetime.strptime(date, '%Y-%m-%d').isocalendar()[1]
df.insert(0, 'date', date)
df.insert(1, 'week', week_of_the_year)
mapping = SPEND_HEADER_MAPPING if goal == "spend" else HEADER_MAPPING
print(df.columns.tolist())
df = df.rename(columns=mapping)
print(df.columns.tolist())
print(df)
df["source_medium"] = df["source_medium"].str.replace(' ', '')
df[["source", "medium"]] = df["source_medium"].str.split('/', expand=True)
df = df.drop(["source_medium"], axis=1)
df["week"] = df["week"].astype(int, copy=False)
return df
def __transform_spend(df: pd.DataFrame) -> pd.DataFrame:
df["spend"] = df["spend"].astype(float, copy=False)
df = __order_columns(df, spend=True)
return df[df.columns[:6]]
def __transform_attribution(df: pd.DataFrame, goal: str) -> pd.DataFrame:
df.insert(2, 'goal', goal)
df["last_non_direct_click_conversions"] = df["last_non_direct_click_conversions"].astype(int, copy=False)
df["last_click_prio_conversions"] = df["last_click_prio_conversions"].astype(int, copy=False)
df["dda_conversions"] = df["dda_conversions"].astype(float, copy=False)
return __order_columns(df)
def transform(df: pd.DataFrame, file_name) -> pd.DataFrame:
goal, date, *_ = file_name.split('_')
df = __common_transformation(df, date, goal)
# we only add goal in attribution df (google-analytics table).
return __transform_spend(df) if "spend" in file_name else __transform_attribution(df, goal)
def main(event, context):
"""Triggered by a change to a Cloud Storage bucket.
Args:
event (dict): Event payload.
context (google.cloud.functions.Context): Metadata for the event.
"""
file = event
file_name = download_from_gcs(file)
df = pd.read_csv(f"/tmp/{file_name}")
transformed_df = transform(df, file_name)
with open(f"/tmp/bq_{file_name}", "w") as file_object:
file_object.write(transformed_df.to_csv(index=False))
with open(f"/tmp/bq_{file_name}", "rb") as file_object:
table = "google-analytics-spend" if "spend" in file_name else "google-analytics"
load_in_bigquery(file_object, dataset='attribution', table=table)
回答1:
You might would prefer to check this thread:
BigQuery displaying wrong results - Duplicating data from Cloud Function?
Very shortly - the function is to be idempotent, and the state of the process (if the data/file was uploaded into BQ or not) should be kept outside of the cloud function. A text file (in some GCS bucket, not inside the cloud function memory, which can be erased as soon as the cloud function execution is finished) is an option, but GCS has plenty of drawbacks in this particular case. For example, a firestore - is much, much better choice.
You might consider the following algorithm -
When you cloud function starts, it should calculate some hash code based on input data - file/object metadata or file/object data or combination of both. That hash - should be unique for the given set of data.
Your cloud function connects to a predefined firestore collection (the project and the name can be provided in the environment variables) and checks if there a document/record with the given hash as an id - already exists or not.
If that hash already exists (the document exists) in the firestore collection - the cloud function finishes its execution and does not do anything else (can do logging, add some additional details into the firestore document if required, etc.). Thus simply finishes its execution.
If that hash is not found (the document does not exist) - the cloud function creates a new document with the given hash as an id. Some metadata details can be added into that document if needed.
Upon the document is created the cloud function continues the main 'workflow'.
A few things to bear in mind.
1/ IAM permissions - the service account under which the cloud function is running - should have relevant permissions on the firestore. Obviously the firestore API is to be enabled in the given project...
2/ What will happen if the cloud function creates a new firestore document, but then failed to load the data into BigQuery (for any reason). It may be that just a check on the firestore document existence is not enough. Thus, a proper 'state' is to be maintained in the firestore document. For example, when a new document is created (in the firestore), there should be a field __state and a value (for example) IN_PROGRESS is assigned to it. Then, when the data is loaded, the cloud function comes back to the firestore and updates that field with the value DONE (for example). But even that is not enough. As you have a load job - there may be cases, when the load is actually successful, but he cloud function failed (any reason including timeout). You might would like to think what to do in that case as well. In any case, having some 'state' monitoring in the firestore may help to understand/investigate the situation with the loading process. Automation of the monitoring might need developing a separate cloud function, but this is a separate story.
3/ As I mentioned in the thread I pointed above (see reasoning in that answer), loading data from inside the cloud function memory is a risky idea. I would suggest to think about that part of your algorithm again.
4/ It might be a good idea to move the loaded file/object from the "input" bucket to some "processed" (or "archive") bucket in case of success, and to move it into a "failure" bucket, in case the load failed. That is to be done in the cloud function code. Failure outcome can also be reflected in the firestore document (i.e. set the value of the __state field to FAILURE).
来源：https://stackoverflow.com/questions/66035917/cloud-function-running-multiple-times-instead-of-once