无法从RStudio连接到sparklyr

Question

我正在尝试从RStudio连接到spark。目前我们使用的是Cloudera Hadoop发行版，其中运行的是Spark (2.2)。我测试了从边缘节点开始的所有东西，我能够创建Spark上下文并执行我的查询。从RStudio到昨天一切都很好，突然我们遇到了RStudio的问题。

library(dplyr)    
library(sparklyr)   
config <- spark_config()    
config$spark.driver.memory <- "8G"    
config$spark.executor.memory <- "8G"    
config$spark.executor.executor <- "2"    
config$spark.executor.cores <- "4"    
config$spark.kryoserializer.buffer.max <- "2000m"    
config$spark.driver.maxResultSize <- "4G"    
config$spark.akka.frameSize <- "768"    
sc <- spark_connect(master="yarn-client",     
version="2.2.0",    
 config=config,     
spark_home = '/opt/cloudera/parcels/SPARK2-2.2.0.cloudera1-1.cdh5.12.0.p0.142354/lib/spark2')

Error in force(code) : 
  Failed while connecting to sparklyr to port (8880) for sessionid (14727): Sparklyr gateway did not respond while retrieving ports information after 60 seconds
    Path: /opt/cloudera/parcels/SPARK2-2.2.0.cloudera1-1.cdh5.12.0.p0.142354/lib/spark2/bin/spark-submit
    Parameters: --class, sparklyr.Shell, '/usr/lib64/R/library/sparklyr/java/sparklyr-2.2-2.11.jar', 8880, 14727
    Log: /tmp/RtmpoNJQEH/file151b437c0313b_spark.log


---- Output Log ----
18/11/12 13:54:50 INFO sparklyr: Session (14727) is starting under 127.0.0.1 port 8880
18/11/12 13:54:50 INFO sparklyr: Session (14727) found port 8880 is not available
18/11/12 13:54:50 INFO sparklyr: Backend (14727) found port 8884 is available
18/11/12 13:54:50 INFO sparklyr: Backend (14727) is registering session in gateway
18/11/12 13:54:50 INFO sparklyr: Backend (14727) is waiting for registration in gateway

---- Error Log ----

我也验证了sparklyr的版本，是0.9.2

能告诉我可能是哪里出了问题吗？

Answer 1

你能试一下吗？

library(httr)
library(sparklyr)
Sys.setenv(SPARK_HOME = '/opt/cloudera/parcels/SPARK2/lib/spark2')
Sys.setenv(YARN_CONF_DIR = '/opt/cloudera/parcels/SPARK2/lib/spark2/conf/yarn-conf/')
config <- list()
config=c(config,list("sparklyr.shell.deploy-mode"="client"))
httr::with_config(
  config = httr::authenticate(user=":", password="", type="gssnegotiate"),
sc <- spark_connect(master = "yarn-client", version = "2.2.0", config = config))
sc

如果启用了SSL和Kerberos，则可能需要使用此选项

library(httr)
library(sparklyr)
set_config(config(cainfo = "/opt/cloudera/security/global_cacerts.pem"))
Sys.setenv(SPARK_HOME = '/opt/cloudera/parcels/SPARK2/lib/spark2')
Sys.setenv(YARN_CONF_DIR = '/opt/cloudera/parcels/SPARK2/lib/spark2/conf/yarn-conf/')
config <- list()
config=c(config,list("sparklyr.shell.keytab"="/PATH/PATH.keytab",
"sparklyr.shell.principal"="user@DOMAIN.COM",
"sparklyr.shell.deploy-mode"="client"
))
httr::with_config(
  config = httr::authenticate(user=":", password="", type="gssnegotiate"),
sc <- spark_connect(master = "yarn-client", version = "2.2.0", config = config))
sc

注意:使用pem格式的根CA路径替换cainfo，在sparklyr.shell.keytab中指定用户密钥表，在sparklyr.shell.principal中指定UPN(用户主体名称