Geoip2's python library doesn't work in pySpark's map function

Question

I'm using geoip2's python library and pySpark to get the geographical address of some IPs. My code is like:

geoDBpath = 'somePath/geoDB/GeoLite2-City.mmdb'
geoPath = os.path.join(geoDBpath)
sc.addFile(geoPath)
reader = geoip2.database.Reader(SparkFiles.get(geoPath))
def ip2city(ip):
    try:
        city = reader.city(ip).city.name
    except:
        city = 'not found'
    return city

I tried

print ip2city("128.101.101.101")

It works. But when I tried to do this in rdd.map:

rdd = sc.parallelize([ip1, ip2, ip3, ip3, ...])
print rdd.map(lambda x: ip2city(x))

It reported

    Traceback (most recent call last):
  File "/home/worker/software/spark/python/pyspark/rdd.py", line 1299, in take
    res = self.context.runJob(self, takeUpToNumLeft, p)
  File "/home/worker/software/spark/python/pyspark/context.py", line 916, in runJob
    port = self._jvm.PythonRDD.runJob(self._jsc.sc(), mappedRDD._jrdd, partitions)
  File "/home/worker/software/spark/python/lib/py4j-0.8.2.1-src.zip/py4j/java_gateway.py", line 538, in __call__
  File "/home/worker/software/spark/python/lib/py4j-0.8.2.1-src.zip/py4j/protocol.py", line 300, in get_return_value
py4j.protocol.Py4JJavaError: An error occurred while calling z:org.apache.spark.api.python.PythonRDD.runJob.
: org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in stage 1.0 failed 1 times, most recent failure: Lost task 0.0 in stage 1.0 (TID 1, localhost): org.apache.spark.api.python.PythonException: Traceback (most recent call last):
  File "/home/worker/software/spark/python/lib/pyspark.zip/pyspark/worker.py", line 98, in main
    command = pickleSer._read_with_length(infile)
  File "/home/worker/software/spark/python/lib/pyspark.zip/pyspark/serializers.py", line 164, in _read_with_length
    return self.loads(obj)
  File "/home/worker/software/spark/python/lib/pyspark.zip/pyspark/serializers.py", line 422, in loads
    return pickle.loads(obj)
TypeError: Required argument 'fileno' (pos 1) not found

    at org.apache.spark.api.python.PythonRunner$$anon$1.read(PythonRDD.scala:166)
    at org.apache.spark.api.python.PythonRunner$$anon$1.<init>(PythonRDD.scala:207)
    at org.apache.spark.api.python.PythonRunner.compute(PythonRDD.scala:125)
    at org.apache.spark.api.python.PythonRDD.compute(PythonRDD.scala:70)
    at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:297)
    at org.apache.spark.rdd.RDD.iterator(RDD.scala:264)
    at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66)
    at org.apache.spark.scheduler.Task.run(Task.scala:88)
    at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
    at java.lang.Thread.run(Thread.java:745)

Could any one tell my how to make the ip2city function work in rdd.map(). Thanks!

Internuncial

Answer 1

It looks like the problem with your code comes from a reader object. It cannot be correctly serialized as a part of a closure and send to the workers.To deal with this you have instantiate it on the workers. One way you can handle this is to use mapPartitions:

from pyspark import SparkFiles

geoDBpath = 'GeoLite2-City.mmdb'
sc.addFile(geoDBpath)

def partitionIp2city(iter):
    from geoip2 import database

    def ip2city(ip):
        try:
           city = reader.city(ip).city.name
        except:
            city = 'not found'
        return city

    reader = database.Reader(SparkFiles.get(geoDBpath))
    return [ip2city(ip) for ip in iter]

rdd = sc.parallelize(['128.101.101.101', '85.25.43.84'])
rdd.mapPartitions(partitionIp2city).collect()

## ['Minneapolis', None]

Whirl

what does this mapPartitions mean? Will it instantiate a new database.Reader for every batch? – Precess

@loganecolss Once per partition. – Whirl

Answer 2

The example from zero323 works. Below is a change to create a loop for each partition of the RDD and then demonstrate the loop structure. It also utilizes the yield to return the results into a dataframe.

from pyspark import SparkFiles

geoDBpath = 'GeoLite2-City.mmdb'
sc.addFile(geoDBpath)
    
def maxmind_ip(ip):
    import geoip2.database
    reader = geoip2.database.Reader(SparkFiles.get(geoDBpath))
    for row in ip:
        try:
            response = reader.city(row.ipaddress)
            ip_lat = str(response.location.latitude)
            ip_long = str(response.location.longitude)
        except:
            #print('Unable to find lat/long for '+ip)
            ip_lat = 'NA'
            ip_long = 'NA'
        #return t.Row('IP_LAT', 'IP_LONG')(ip_lat, ip_long)
        yield [row.ipaddress, ip_lat, ip_long]
    reader.close()
    
ip_maxmind_results = df_actIP_small.rdd.mapPartitions(maxmind_ip).toDF(["ipaddress","IP_LAT","IP_LONG"]) 

Ruff