Could not bind on a random free port error while trying to connect to spark master

13
votes

I have a spark master running on amazon ec2. I tried to connect to it using pyspark as follows from another ec2 instance as follows:

spark = SparkSession.builder.appName("MyApp") \
                            .master("spark_url_as_obtained_in_web_ui") \
                            .getOrCreate()

The following were the errors:

To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).

2018-04-04 20:03:04 WARN Utils:66 - Service 'sparkDriver' could not bind on a random free port. You may check whether configuring an appropriate binding address.

............

java.net.BindException: Cannot assign requested address: Service 'sparkDriver' failed after 16 retries (on a random free port)! Consider explicitly setting the appropriate binding address for the service 'sparkDriver' (for example spark.driver.bindAddress for SparkDriver) to the correct binding address.

I tried all the solutions as described here but to no avail:

  1. Connecting to a remote Spark master - Java / Scala

  2. All masters are unresponsive ! ? Spark master is not responding with datastax architecture

  3. Spark Standalone Cluster - Slave not connecting to Master

  4. Spark master-machine:7077 not reachable

  5. spark submit "Service 'Driver' could not bind on port" error

  6. https://community.hortonworks.com/questions/8257/how-can-i-resolve-it.html

What could be going wrong??

3
python-3.xapache-sparkamazon-ec2pyspark

3 Answers

16
votes

Set spark.driver.bindAddress to your local IP like 127.0.0.1.

pyspark -c spark.driver.bindAddress=127.0.0.1
7
votes

While creating spark session set the below configurations

spark =SparkSession.builder.appName(str(name))\
      .master("local[*]").config("spark.driver.memory","5g")\
      .config("spark.driver.host","10.51.4.110")\ #Machine ip
      .config("spark.driver.bindAddress","10.51.4.110")\ #Machine ip
      .getOrCreate()

Sometimes In addition to bind address, we need to set the host address also. In my case system host address changed to system name. Spark shows timeout error. then after setting host and bind address as same. It works fine

*10.51.4.110 - Local Machine IP

3
votes

I had a similar issue on Windows 10 recently.

Resolved it by following below steps:

  • Fixed it by setting a user environment variable SPARK_LOCAL_IP = 127.0.0.1
  • Restart the command line as an admin