-
Notifications
You must be signed in to change notification settings - Fork 0
/
spark_streaming.py
171 lines (146 loc) · 6.23 KB
/
spark_streaming.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
import logging
import logging.config
from configparser import ConfigParser
from pyspark.sql import SparkSession
from pyspark.sql.types import *
import pyspark.sql.functions as psf
import subprocess
import os
def run_cmd(args_list):
"""
run linux commands
"""
# import subprocess
# print('Running system command: {0}'.format(' '.join(args_list)))
proc = subprocess.Popen(args_list, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
s_output, s_err = proc.communicate()
s_return = proc.returncode
if s_return != 0:
return s_return, s_output, s_err
else:
return "Command executed successfully!"
def insert_batch(df, epoch_id, target_table):
df = (
df.withColumn("part_date", psf.from_unixtime(psf.col("ts")/1000, "YYYYMMdd"))
.select("ts", "userid", "sessionid", "page", "iteminsession", "auth", "method", "status", "level", "location", "useragent", "lastname",
"firstname", "registration", "gender", "artist", "song", "length", "part_date")
)
df.coalesce(1).write.mode("append").insertInto(target_table, overwrite=False)
# df.coalesce(1).write.partitionBy("part_date").format("parquet").mode("append").saveAsTable(target_table)
run_cmd(["impala-shell", "-i", "bdp-worker01-pdc.vn.prod", "-k", "--ssl", "-q", f"REFRESH {target_table}"])
def run_spark_job(spark: SparkSession, config: ConfigParser):
"""
Run Spark Structured Streaming job reading data from Kafka
"""
# set log level for Spark app
spark.sparkContext.setLogLevel("WARN")
# define schema for incoming data
kafka_schema = StructType([
StructField("ts", LongType(), True),
StructField("userId", StringType(), True),
StructField("sessionId", IntegerType(), True),
StructField("page", StringType(), True),
StructField("itemInSession", IntegerType(), True),
StructField("auth", StringType(), True),
StructField("method", StringType(), True),
StructField("status", StringType(), True),
StructField("level", StringType(), True),
StructField("location", StringType(), True),
StructField("userAgent", StringType(), True),
StructField("lastName", StringType(), True),
StructField("firstName", StringType(), True),
StructField("registration", LongType(), True),
StructField("gender", StringType(), True),
StructField("artist", StringType(), True),
StructField("song", StringType(), True),
StructField("length", DoubleType(), True)
])
# start reading data from Kafka
df = spark \
.readStream \
.format("kafka") \
.option("kafka.bootstrap.servers", config.get("spark", "bootstrap_servers")) \
.option("subscribe", config.get("kafka", "topic")) \
.option("startingOffsets", config.get("spark", "starting_offsets")) \
.option("maxOffsetsPerTrigger", config.get("spark", "max_offsets_per_trigger")) \
.option("maxRatePerPartition", config.get("spark", "max_rate_per_partition")) \
.option("stopGracefullyOnShutdown", "true") \
.load()
# print schema of incoming data
logging.debug("Printing schema of incoming data")
df.printSchema()
# extract value of incoming Kafka data, ignore key
kafka_df = df.selectExpr("CAST(value AS STRING)")
service_table = kafka_df \
.select(psf.from_json(kafka_df.value, kafka_schema).alias("DF")) \
.select("DF.*")
# query = service_table.writeStream.trigger(processingTime="20 seconds").format("console").option("truncate", "false").start()
service_table.writeStream.format("console").option("truncate", "false").start()
target_table = config.get("spark", "target_table")
if config.get("spark", "checkpoint_remove") == "True":
run_cmd(["hdfs", "dfs", "-rm", "-r", config.get("spark", "checkpoint_dir")])
query = (service_table
.writeStream
.outputMode("append")
.option("checkpointLocation", config.get("spark", "checkpoint_dir"))
.foreachBatch(lambda df, epochId: insert_batch(df, epochId, target_table)
).start())
query.awaitTermination()
if __name__ == "__main__":
cur_path = os.getcwd()
# load config
config = ConfigParser()
config.read(os.path.join(cur_path, "app.cfg"))
# start logging
logging.config.fileConfig(os.path.join(cur_path, "logging.ini"))
logger = logging.getLogger(__name__)
# create spark session
spark = (
SparkSession
.builder
.master(config.get("spark", "master"))
.appName("itbi.streaming.consumer.test")
.config("spark.sql.sources.partitionOverwriteMode","dynamic")
.config("hive.exec.dynamic.partition", "true")
.config("hive.exec.dynamic.partition.mode", "nonstrict")
.config("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
.config("spark.hadoop.hive.exec.stagingdir", "/tmp/simple_kompactor")
.getOrCreate()
)
target_table = config.get("spark", "target_table")
if config.get("spark", "drop_table") == "True":
pass
# Prepare output table
spark.sql(f"DROP TABLE IF EXISTS {target_table}")
create_table_query = f"""
CREATE TABLE IF NOT EXISTS {target_table}
(
`ts` bigint,
`userid` string,
`sessionid` int,
`page` string,
`iteminsession` int,
`auth` string,
`method` string,
`status` string,
`level` string,
`location` string,
`useragent` string,
`lastname` string,
`firstname` string,
`registration` bigint,
`gender` string,
`artist` string,
`song` string,
`length` double
)
COMMENT 'Testing table with Kafka'
PARTITIONED BY (`part_date` string)
STORED AS PARQUET
TBLPROPERTIES("auto.purge"="true")
"""
spark.sql(create_table_query)
logger.info("Starting Spark Job")
run_spark_job(spark, config)
logger.info("Closing Spark Session")
spark.stop()