Dataflow (Apache Beam) 無法寫入 BigQuery

Question

我有一個管道，必須在其最后步驟中在 BigQuery 上寫入兩條記錄，我真的不知道為什么它似乎沒有插入任何內容。 我沒有錯誤，該表存在並且它已經包含記錄，實際上我必須使用 TRUNCATE/INSERT 模式。

有人可以幫我弄清楚為什么它沒有像我預期的那樣工作嗎？

這是我的管道：

     p = beam.Pipeline(options=pipeline_options)

    (p
        | 'Read Configuration Table ' >> beam.io.Read(beam.io.BigQuerySource(config['ENVIRONMENT']['configuration_table']))
        | 'Get Files from Server' >> beam.Map(import_file)
        | 'Upload files on Bucket' >> beam.Map(upload_file_on_bucket)
        | 'Set record update' >> beam.Map(set_last_step)
        | 'Update table' >> beam.io.gcp.bigquery.WriteToBigQuery(
                table=config['ENVIRONMENT']['configuration_table'],
                write_disposition='WRITE_TRUNCATE',
                schema=('folder:STRING, last_file:STRING')
                )
     )

和

def set_last_step(file_list):
    logging.info(msg='UPDATE CONFIGURATION TABLE - working on: ' + str(file_list))
    folder = ''

    if 'original' in file_list:
        if '1951' in file_list:
            folder = '1951'
        else:
            folder = '1952'
        dic = {'folder': folder, 'last_file': file_list['original']}
        logging.info(msg='UPDATE CONFIGURATION TABLE - no work done, reporting original record: ' + str(dic))
    else:
        folder = list(file_list.keys())[0]
        path = list(file_list.values())[0]
        dic = {'folder': folder, 'last_file': path}
        logging.info(msg='UPDATE CONFIGURATION TABLE - work done, reporting new record: ' + str(dic))

    purge(dir=os.path.join(HOME_PATH, 'download'), pattern=folder+"_")

    logging.info(msg='UPDATE CONFIGURATION TABLE - record to be updated: ' + str(dic))

    return dic

WriteToBigQuery 階段的輸入記錄（顯然是來自“更新表”階段的 output）是：

{'folder': '1952', 'last_file': '1952_2019120617.log.gz'}
{'folder': '1951', 'last_file': '1951_2019120617.log.gz'}

來自 DataFlow 的調試信息是：

2019-12-06 18:09:36 DEBUG    Creating or getting table <TableReference
 datasetId: 'MYDATASET'
 projectId: 'MYPROJECT'
 tableId: 'MYTABLE'> with schema {'fields': [{'name': 'folder', 'type': 'STRING', 'mode': 'NULLABLE'}, {'name': 'last_file', 'type': 'STRING', 'mode': 'NULLABLE'}]}.
2019-12-06 18:09:36 DEBUG    Created the table with id MYTABLE
2019-12-06 18:09:36 INFO     Created table MYPROJECT.MYDATASET.MYTABLE with schema <TableSchema
 fields: [<TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'folder'
 type: 'STRING'>, <TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'last_file'
 type: 'STRING'>]>. Result: <Table
 creationTime: 1575652176727
 etag: '0/GXOOeXPCmYsMfgGNxl2Q=='
 id: 'MYPROJECT:MYDATASET.MYTABLE'
 kind: 'bigquery#table'
 lastModifiedTime: 1575652176766
 location: 'EU'
 numBytes: 0
 numLongTermBytes: 0
 numRows: 0
 schema: <TableSchema
 fields: [<TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'folder'
 type: 'STRING'>, <TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'last_file'
 type: 'STRING'>]>
 selfLink: 'https://www.googleapis.com/bigquery/v2/projects/MYPROJECT/datasets/MYDATASET/tables/MYTABLE'
 tableReference: <TableReference
 datasetId: 'MYDATASET'
 projectId: 'MYPROJECT'
 tableId: 'MYTABLE'> with schema {'fields': [{'name': 'folder', 'type': 'STRING', 'mode': 'NULLABLE'}, {'name': 'last_file', 'type': 'STRING', 'mode': 'NULLABLE'}]}.
2019-12-06 18:09:36 DEBUG    Created the table with id MYTABLE
2019-12-06 18:09:36 INFO     Created table MYPROJECT.MYDATASET.MYTABLE with schema <TableSchema
 fields: [<TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'folder'
 type: 'STRING'>, <TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'last_file'
 type: 'STRING'>]>. Result: <Table
 creationTime: 1575652176727
 etag: '0/GXOOeXPCmYsMfgGNxl2Q=='
 id: 'MYPROJECT:MYDATASET.MYTABLE'
 kind: 'bigquery#table'
 lastModifiedTime: 1575652176766
 location: 'EU'
 numBytes: 0
 numLongTermBytes: 0
 numRows: 0
 schema: <TableSchema
 fields: [<TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'folder'
 type: 'STRING'>, <TableFieldSchema
 fields: []
 mode: 'NULLABLE'
 name: 'last_file'
 type: 'STRING'>]>
 selfLink: 'https://www.googleapis.com/bigquery/v2/projects/MYPROJECT/datasets/MYDATASET/tables/MYTABLE'
 tableReference: <TableReference
 datasetId: 'MYDATASET'
 projectId: 'MYPROJECT'
 tableId: 'MYTABLE'>
 type: 'TABLE'>.
2019-12-06 18:09:36 WARNING  Sleeping for 150 seconds before the write as BigQuery inserts can be routed to deleted table for 2 mins after the delete and create.
2019-12-06 18:12:06 DEBUG    Attempting to flush to all destinations. Total buffered: 2
2019-12-06 18:12:06 DEBUG    Flushing data to MYPROJECT:MYDATASET.MYTABLE. Total 2 rows.
2019-12-06 18:12:07 DEBUG    Passed: True. Errors are []

Answer 1

在此示例中，我將 XML 元素解析為 DF 並將其推送到 GBQ。 希望你能在這里找到有用的東西。

import pandas as pd
import xml.etree.ElementTree as ET
import datetime
import json
import requests
import pandas_gbq
from lxml import etree

# authentication: working now....
login = 'FN.LN@your_email.com' 
password = 'your_AS_pswd'


AsOfDate = datetime.datetime.today().strftime('%m-%d-%Y')

#1) SLA=471162: Execute Query
REQUEST_URL = 'https://www.some_data.com'
response = requests.get(REQUEST_URL, auth=(login, password))
xml_data = response.text.encode('utf-8', 'ignore') 

#print(response.text)

#tree = etree.parse(xml_data)
root = ET.fromstring(xml_data)

# start collecting root elements and headers for data frame 1
desc = root.get("SLA_Description")
frm = root.get("start_date")
thru = root.get("end_date")
dev = root.get("obj_device")
loc = root.get("locations")
loc = loc[:-1]
df1 = pd.DataFrame([['From:',frm],['Through:',thru],['Object:',dev],['Location:',loc]])
df1.columns = ['SLAs','Analytics']
#print(df1)

# start getting the analytics for data frame 2
data=[['Goal:',root[0][0].text],['Actual:',root[0][1].text],['Compliant:',root[0][2].text],['Errors:',root[0][3].text],['Checks:',root[0][4].text]]
df2 = pd.DataFrame(data)
df2.columns = ['SLAs','Analytics']
#print(df2)

# merge data frame 1 with data frame 2
df3 = df1.append(df2, ignore_index=True)
#print(df3)

# append description and today's date onto data frame
df3['Description'] = desc
df3['AsOfDate'] = AsOfDate

#df3.dtypes

# push from data frame, where data has been transformed, into Google BQ
pandas_gbq.to_gbq(df3, 'website.Metrics', 'your-firm', chunksize=None, reauth=False, if_exists='append', private_key=None, auth_local_webserver=False, table_schema=None, location=None, progress_bar=True, verbose=None)
print('Execute Query, Done!!')

Answer 2

該問題與 Dataflow 在 BigQuery 上使用 stream 方法這一事實有關，這意味着數據不會立即保存在數據庫中，而是在 window 之后保存...所以您只需等待幾分鍾。