Search code examples
javascalajarapache-sparkcluster-computing

Spark spark-submit --jars arguments wants comma list, how to declare a directory of jars?

In Submitting Applications in the Spark docs, as of 1.6.0 and earlier, it's not clear how to specify the --jars argument, as it's apparently not a colon-separated classpath not a directory expansion.

The docs say "Path to a bundled jar including your application and all dependencies. The URL must be globally visible inside of your cluster, for instance, an hdfs:// path or a file:// path that is present on all nodes."

Question: What are all the options for submitting a classpath with --jars in the spark-submit script in $SPARK_HOME/bin? Anything undocumented that could be submitted as an improvement for docs?

I ask because when I was testing --jars today, we had to explicitly provide a path to each jar:

/usr/local/spark/bin/spark-submit --class jpsgcs.thold.PipeLinkageData ---jars=local:/usr/local/spark/jars/groovy-all-2.3.3.jar,local:/usr/local/spark/jars/guava-14.0.1.jar,local:/usr/local/spark/jars/jopt-simple-4.6.jar,local:/usr/local/spark/jars/jpsgcs-core-1.0.8-2.jar,local:/usr/local/spark/jars/jpsgcs-pipe-1.0.6-7.jar /usr/local/spark/jars/thold-0.0.1-1.jar

We are choosing to pre-populate the cluster with all the jars in /usr/local/spark/jars on each worker, it seemed that if no local:/ file:/ or hdfs: was supplied, then the default is file:/ and the driver makes the jars available on a webserver run by the driver. I chose local, as above.

And it seems that we do not need to put the main jar in the --jars argument, I have not tested yet if other classes in the final argument (application-jar arg per docs, i.e. /usr/local/spark/jars/thold-0.0.1-1.jar) are shipped to workers, or if I need to put the application-jar in the --jars path to get classes not named after --class to be seen.

(And granted with Spark standalone mode using --deploy-mode client, you also have to put a copy of the driver on each worker but you don't know up front which worker will run the driver)

Solution

  • In this way it worked easily.. instead of specifying each jar with version separately..

    #!/bin/sh
# build all other dependent jars in OTHER_JARS

JARS=`find ../lib -name '*.jar'`
OTHER_JARS=""
   for eachjarinlib in $JARS ; do    
if [ "$eachjarinlib" != "APPLICATIONJARTOBEADDEDSEPERATELY.JAR" ]; then
       OTHER_JARS=$eachjarinlib,$OTHER_JARS
fi
done
echo ---final list of jars are : $OTHER_JARS
echo $CLASSPATH

spark-submit --verbose --class <yourclass>
... OTHER OPTIONS
--jars $OTHER_JARS,APPLICATIONJARTOBEADDEDSEPERATELY.JAR

    • Using tr unix command also can help like the below example.

      --jars $(echo /dir_of_jars/*.jar | tr ' ' ',')

    • One parent directory and recrsive directorires (each dir have different jars) ... case

    $(find /dir_of_jars -name '*.jar' -print0 | tr '\0' ',' | sed 's/,$//')