之前用的版本是zeppelin0.7.2和spark1.6,最近把zeppelin升级到0.8发现1.6的配置和2.2的配置默认都无法支持,大概会遇到3个问题。
Incompatible Jackson version: 2.8.11-1
jackson版本问题
rm zeppelin/lib/jackson-*
cp $spark_home/jars/jackson-* zeppelin/lib/
就可以解决了org.apache.spark.network.client.TransportResponseHandler.channelInactive
这个是因为netty包版本不一致问题导致
同样
rm zeppelin/lib/netty*
cp $spark_home/jars/netty* zeppelin/lib
搞定了spark版本未识别
在zeppelin interpreter 中,配置SPARK_MAJOR_VERSION 值为2
其他参数如果需要可以参照,按照个人喜好来吧
Properties
name value
SPARK_HOME /data/sysdir/spark2.2
SPARK_MAJOR_VERSION 2
args
master yarn-client
spark.app.name Zeppelin
spark.cores.max 3
spark.driver.extraJavaOptions -XX:PermSize=2g -XX:MaxPermSize=2g
spark.driver.maxResultSize 8g
spark.dynamicAllocation.enabled false
spark.executor.instances 150
spark.executor.memory 8g
spark.kryoserializer.buffer.max 2000m
spark.yarn.executor.memoryOverhead 4086
spark.yarn.queue xxx.xxx (yarn队列名称)
yarn.nodemanager.pmem-check-enabled false
yarn.nodemanager.vmem-check-enabled false
zeppelin.R.cmd R
zeppelin.R.image.width 100%
zeppelin.R.knitr true
zeppelin.R.render.options out.format = ‘html’, comment = NA, echo = FALSE, results = ‘asis’, message = F, warning = F, fig.retina = 2
zeppelin.dep.additionalRemoteRepository spark-packages,http://dl.bintray.com/spark-packages/maven,false;
zeppelin.dep.localrepo local-repo
zeppelin.interpreter.localRepo /data/zeppelin-0.8.0-bin-all/local-repo/3CM5P9NU9
zeppelin.interpreter.output.limit 10240000
zeppelin.pyspark.python java
zeppelin.pyspark.useIPython true
zeppelin.spark.concurrentSQL false
zeppelin.spark.enableSupportedVersionCheck true
zeppelin.spark.importImplicit true
zeppelin.spark.maxResult 10000
zeppelin.spark.printREPLOutput true
zeppelin.spark.sql.interpolation true
zeppelin.spark.sql.stacktrace true
zeppelin.spark.uiWebUrl
zeppelin.spark.useHiveContext true
zeppelin.spark.useNew true