I believe this is due to SparkInterpreterLauncher doesn't support parameters with whitespace. (It would use whitespace as delimiter to separate parameters), this is a known issue
Lior Chaga <lio...@taboola.com> 于2021年7月11日周日 下午4:14写道: > So after adding the quotes in both SparkInterpreterLauncher > and interpreter.sh, interpreter is still failing with same error of > Unrecognized option. > But the weird thing is that if I copy the command supposedly executed from > zeppelin (as it is printed to log) and run it directly in shell, the > interpreter process is properly running. So my guess is that the forked > process command that is created, is not really identical to the one that is > logged. > > This is how my cmd looks like (censored a bit): > > /usr/local/spark/bin/spark-submit > --class org.apache.zeppelin.interpreter.remote.RemoteInterpreterServer > --driver-class-path > :/zeppelin/local-repo/spark/*:/zeppelin/interpreter/spark/*:::/zeppelin/inter > preter/zeppelin-interpreter-shaded-0.10.0-SNAPSHOT.jar:/zeppelin/interpreter/spark/spark-interpreter-0.10.0-SNAPSHOT.jar:/etc/hadoop/conf > > *--driver-java-options " -DSERVICENAME=zeppelin_docker > -Dfile.encoding=UTF-8 > -Dlog4j.configuration=file:///zeppelin/conf/log4j.properties > -Dlog4j.configurationFile=file:///zeppelin/conf/log4j2.properties > -Dzeppelin.log.file=/var/log/zeppelin/zeppelin-interpreter-spark-shared_process--zeppelin-test-spark3-7d74d5df4-2g8x5.log" > * > --conf spark.driver.host=10.135.120.245 > --conf "spark.dynamicAllocation.minExecutors=1" > --conf "spark.shuffle.service.enabled=true" > --conf "spark.sql.parquet.int96AsTimestamp=true" > --conf "spark.ui.retainedTasks=10000" > --conf "spark.executor.heartbeatInterval=600s" > --conf "spark.ui.retainedJobs=100" > --conf "spark.sql.ui.retainedExecutions=10" > --conf "spark.hadoop.cloneConf=true" > --conf "spark.debug.maxToStringFields=200000" > --conf "spark.executor.memory=70g" > --conf > "spark.executor.extraClassPath=../mysql-connector-java-8.0.18.jar:../guava-19.0.jar" > > --conf "spark.hadoop.fs.permissions.umask-mode=000" > --conf "spark.memory.storageFraction=0.1" > --conf "spark.scheduler.mode=FAIR" > --conf "spark.sql.adaptive.enabled=true" > --conf > "spark.master=mesos://zk://zk003:2181,zk004:2181,zk006:2181,/mesos-zeppelin" > > --conf "spark.driver.memory=15g" > --conf "spark.io.compression.codec=lz4" > --conf "spark.executor.uri= > https://artifactory.company.com/artifactory/static/spark/spark-dist/spark-3.1.2.2-hadoop-2.7-zulu" > - > -conf "spark.ui.retainedStages=500" > --conf "spark.mesos.uris= > https://artifactory.company.com/artifactory/static/spark/spark-executor/jars/mysql-connector-java-8.0.18.jar,https://artifactory.company.com/artifactory/static/spark/spark-executor/jars/guava-19.0.jar" > > --conf "spark.driver.maxResultSize=8g" > *--conf "spark.executor.extraJavaOptions=-DSERVICENAME=Zeppelin > -agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=2015 > -XX:-OmitStackTraceInFastThrow -Dcom.sun.management.jmxremote > -Dcom.sun.management.jmxremote.port=55745 > -Dcom.sun.management.jmxremote.authenticate=false > -Dcom.sun.management.jmxremote.ssl=false -verbose:gc > -Dlog4j.configurationFile=/etc/config/log4j2-executor-config.xml > -XX:+UseG1GC -verbose:gc -XX:+PrintHeapAtGC -XX:+PrintGCDateStamps > -XX:+PrintGCTimeStamps -XX:+PrintFlagsFinal -XX:+PrintReferenceGC > -XX:+PrintGCDetails -XX:+PrintAdaptiveSizePolicy > -XX:+UnlockDiagnosticVMOptions -XX:+G1SummarizeConcMark > -XX:+PrintStringDeduplicationStatistics -XX:+UseStringDeduplication > -XX:InitiatingHeapOccupancyPercent=35 > -Dhttps.proxyHost=proxy.service.consul -Dhttps.proxyPort=3128" * > --conf "spark.dynamicAllocation.enabled=true" > --conf "spark.default.parallelism=1200" > --conf "spark.hadoop.mapreduce.fileoutputcommitter.algorithm.version=2" > --conf > "spark.hadoop.fs.AbstractFileSystem.gs.impl=com.google.cloud.hadoop.fs.gcs.GoogleHadoopFS" > > --conf "spark.app.name=zeppelin_docker_spark3" > --conf "spark.shuffle.service.port=7337" > --conf "spark.memory.fraction=0.75" > --conf "spark.mesos.coarse=true" > --conf "spark.ui.port=4041" > --conf "spark.dynamicAllocation.executorIdleTimeout=60s" > --conf "spark.sql.shuffle.partitions=1200" > --conf "spark.sql.parquet.outputTimestampType=TIMESTAMP_MILLIS" > --conf "spark.dynamicAllocation.cachedExecutorIdleTimeout=120s" > --conf "spark.network.timeout=1200s" > --conf "spark.cores.max=600" > --conf > "spark.hadoop.fs.gs.impl=com.google.cloud.hadoop.fs.gcs.GoogleHadoopFileSystem" > > --conf "spark.worker.timeout=150000" > *--conf > "spark.driver.extraJavaOptions=-Dhttps.proxyHost=proxy.service.consul > -Dhttps.proxyPort=3128 > -Dlog4j.configuration=file:/usr/local/spark/conf/log4j.properties > -Djavax.jdo.option.ConnectionDriverName=com.mysql.cj.jdbc.Driver > -Djavax.jdo.option.ConnectionPassword=2eebb22277 > -Djavax.jdo.option.ConnectionURL=jdbc:mysql://proxysql-backend.service.consul.company.com:6033/hms?useSSL=false&databaseTerm=SCHEMA&nullDatabaseMeansCurrent=true > <http://proxysql-backend.service.consul.company.com:6033/hms?useSSL=false&databaseTerm=SCHEMA&nullDatabaseMeansCurrent=true> > -Djavax.jdo.option.ConnectionUserName=hms_rw" * > --conf "spark.files.overwrite=true" > /zeppelin/interpreter/spark/spark-interpreter-0.10.0-SNAPSHOT.jar > 10.135.120.245 > 36419 > spark-shared_process : > > > > *Error: Unrecognized option: > -agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=2015* > > Will continue tackling it... > > On Thu, Jul 8, 2021 at 4:49 PM Jeff Zhang <zjf...@gmail.com> wrote: > >> Thanks Lior for the investigation. >> >> >> Lior Chaga <lio...@taboola.com> 于2021年7月8日周四 下午8:31写道: >> >>> Ok, I think I found the issue. It's not only that the quotations are >>> missing from the --conf param, they are also missing from >>> the --driver-java-options, which is concatenated to >>> the INTERPRETER_RUN_COMMAND in interpreter.sh >>> >>> I will fix it in my build, but would like a confirmation that this is >>> indeed the issue (and I'm not missing anything), so I'd open a pull >>> request. >>> >>> On Thu, Jul 8, 2021 at 3:05 PM Lior Chaga <lio...@taboola.com> wrote: >>> >>>> I'm trying to run zeppelin using local spark interpreter. >>>> Basically everything works, but if I try to set >>>> `spark.driver.extraJavaOptions` or `spark.executor.extraJavaOptions` >>>> containing several arguments, I get an exception. >>>> For instance, for providing `-DmyParam=1 -DmyOtherParam=2`, I'd get: >>>> Error: Unrecognized option: -DmyOtherParam=2 >>>> >>>> I noticed that the spark submit looks as follow: >>>> >>>> spark-submit --class >>>> org.apache.zeppelin.interpreter.remote.RemoteInterpreterServer >>>> --driver-class-path >>>> .... *--conf spark.driver.extraJavaOptions=-DmyParam=1 >>>> -DmyOtherParam=2* >>>> >>>> So I tried to patch SparkInterpreterLauncher to add quotation marks >>>> (like in the example from spark documentation - >>>> https://spark.apache.org/docs/latest/configuration.html#dynamically-loading-spark-properties >>>> ) >>>> >>>> I see that the quotation marks were added: *--conf >>>> "spark.driver.extraJavaOptions=-DmyParam=1 -DmyOtherParam=2"* >>>> But I still get the same error. >>>> >>>> Any idea how I can make it work? >>>> >>> >> >> -- >> Best Regards >> >> Jeff Zhang >> > -- Best Regards Jeff Zhang