Sparking Water Shell: Cloud size under 12 Exception

In my last blog, I compared Sparking Water and H2O. Before I made Sparking-shell work, I run into a lot of issues. One of annoying errors was runtime exception: Cloud size under xx. Searched internet and found many people have the similar problems. There are many recommendations, ranging from downloading the latest and matching version, to set to certain parameters during startup. Unfortunately none of them were working for me. But finally I figured out the issue and would like to share my solution in this blog.

After I downloaded Sparking Water, unzipped the file, and run sparking-shell command as shown from It looked good initially.

[sparkling-water-2.2.2]$ bin/sparkling-shell --conf "spark.executor.memory=1g"

  Spark master (MASTER)     : local[*]
  Spark home   (SPARK_HOME) : /opt/cloudera/parcels/SPARK2-2.2.0.cloudera1-1.cdh5.12.0.p0.142354/lib/spark2
  H2O build version         : (weierstrass)
  Spark build version       : 2.2.0
  Scala version             : 2.11

Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
Spark context Web UI available at
Spark context available as 'sc' (master = yarn, app id = application_3608740912046_1343).
Spark session available as 'spark'.
Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /___/ .__/\_,_/_/ /_/\_\   version 2.2.0.cloudera1

Using Scala version 2.11.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_144)
Type in expressions to have them evaluated.
Type :help for more information.


But when I run command val h2oContext = H2OContext.getOrCreate(spark), it gave me many errors as follows:

scala> import org.apache.spark.h2o._
import org.apache.spark.h2o._

scala> val h2oContext = H2OContext.getOrCreate(spark)
17/11/05 10:07:48 WARN internal.InternalH2OBackend: Increasing 'spark.locality.wait' to value 30000
17/11/05 10:07:48 WARN internal.InternalH2OBackend: Due to non-deterministic behavior of Spark broadcast-based joins
We recommend to disable them by configuring `spark.sql.autoBroadcastJoinThreshold` variable to value `-1`:
sqlContext.sql("SET spark.sql.autoBroadcastJoinThreshold=-1")
17/11/05 10:07:48 WARN internal.InternalH2OBackend: The property 'spark.scheduler.minRegisteredResourcesRatio' is not specified!
We recommend to pass `--conf spark.scheduler.minRegisteredResourcesRatio=1`
17/11/05 10:07:48 WARN internal.InternalH2OBackend: Unsupported options spark.dynamicAllocation.enabled detected!
17/11/05 10:07:48 WARN internal.InternalH2OBackend:
The application is going down, since the parameter (,true) is true!
If you would like to skip the fail call, please, specify the value of the parameter to false.

java.lang.IllegalArgumentException: Unsupported argument: (spark.dynamicAllocation.enabled,true)
  at org.apache.spark.h2o.backends.internal.InternalBackendUtils$$anonfun$checkUnsupportedSparkOptions$1.apply(InternalBackendUtils.scala:46)
  at org.apache.spark.h2o.backends.internal.InternalBackendUtils$$anonfun$checkUnsupportedSparkOptions$1.apply(InternalBackendUtils.scala:38)
  at scala.collection.immutable.List.foreach(List.scala:381)
  at org.apache.spark.h2o.backends.internal.InternalBackendUtils$class.checkUnsupportedSparkOptions(InternalBackendUtils.scala:38)
  at org.apache.spark.h2o.backends.internal.InternalH2OBackend.checkUnsupportedSparkOptions(InternalH2OBackend.scala:30)
  at org.apache.spark.h2o.backends.internal.InternalH2OBackend.checkAndUpdateConf(InternalH2OBackend.scala:60)
  at org.apache.spark.h2o.H2OContext.<init>(H2OContext.scala:90)
  at org.apache.spark.h2o.H2OContext$.getOrCreate(H2OContext.scala:355)
  at org.apache.spark.h2o.H2OContext$.getOrCreate(H2OContext.scala:383)
  ... 50 elided

You can see I need to pass in more parameters when starting sparking-shell. Change the parameters as follows:

bin/sparkling-shell \
--master yarn \
--conf spark.executor.memory=1g \
--conf spark.scheduler.maxRegisteredResourcesWaitingTime=1000000 \
--conf \
--conf spark.dynamicAllocation.enabled=false \
--conf spark.sql.autoBroadcastJoinThreshold=-1 \
--conf spark.locality.wait=30000 \
--conf spark.scheduler.minRegisteredResourcesRatio=1

Ok, this time it looked better, at least warning messages disappeared. But got error message java.lang.RuntimeException: Cloud size under 2.

[sparkling-water-2.2.2]$ bin/sparkling-shell \
> --master yarn \
> --conf spark.executor.memory=1g \
> --conf spark.scheduler.maxRegisteredResourcesWaitingTime=1000000 \
> --conf \
> --conf spark.dynamicAllocation.enabled=false \
> --conf spark.sql.autoBroadcastJoinThreshold=-1 \
> --conf spark.locality.wait=30000 \
> --conf spark.scheduler.minRegisteredResourcesRatio=1

  Spark master (MASTER)     : yarn
  Spark home   (SPARK_HOME) : /opt/cloudera/parcels/SPARK2-2.2.0.cloudera1-1.cdh5.12.0.p0.142354/lib/spark2
  H2O build version         : (weierstrass)
  Spark build version       : 2.2.0
  Scala version             : 2.11

Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
Spark context Web UI available at
Spark context available as 'sc' (master = yarn, app id = application_3608740912046_1344).
Spark session available as 'spark'.
Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /___/ .__/\_,_/_/ /_/\_\   version 2.2.0.cloudera1

Using Scala version 2.11.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_144)
Type in expressions to have them evaluated.
Type :help for more information.

scala> import org.apache.spark.h2o._
import org.apache.spark.h2o._

scala> val h2oContext = H2OContext.getOrCreate(spark)
java.lang.RuntimeException: Cloud size under 2
  at water.H2O.waitForCloudSize(
  at org.apache.spark.h2o.backends.internal.InternalH2OBackend.init(InternalH2OBackend.scala:117)
  at org.apache.spark.h2o.H2OContext.init(H2OContext.scala:121)
  at org.apache.spark.h2o.H2OContext$.getOrCreate(H2OContext.scala:355)
  at org.apache.spark.h2o.H2OContext$.getOrCreate(H2OContext.scala:383)
  ... 50 elided

Not big deal. Anyway I didn’t specify the number of executors used and executor memory. It may complain about the size of the H2O cluster is too small. Add the following three parameters.

–conf spark.executor.instances=12 \
–conf spark.executor.memory=10g \
–conf spark.driver.memory=8g \

Rerun the whole thing got the same error with the size number changing to 12. It did not look right to me. Then I check out the H2O error logfile and found tons of messages as follows:

11-06 10:23:16.355  30452  #09:54321 ERRR: Got IO error when sending batch UDP bytes: Connection refused
11-06 10:23:16.790  30452  #06:54321 ERRR: Got IO error when sending batch UDP bytes: Connection refused

It looks like Sparking Water can not connect to Spark cluster. After some investigation, I then realized I installed and run sparking-shell from edge node on the BDA. If the H2O cluster was running inside a Spark application, the communication of Spark cluster on BDA is through BDA’s private network, or InfiniteBand network. Edge node can not directly communicate to IB network on BDA. With this assumption in mind, I installed and run Sparking Water on one of BDA nodes, it worked perfectly without any issue. Problem solved!


H2O vs Sparkling Water

People working in Hadoop environment are familiar with many products that make you feel like you’re in a zoo. For example, Pig, Hive, Beeswax, and ZooKeeper are some of them. As machine learning becomes more popular, products sounds like water came out, such as H2O and Sparking Water. I am not going to rule out the possibility we will see some big data products sound like wine. Anyway, people like call various new names to make their products sound cool, although many of them are similar.

In this blog, I am going to discuss H2O and Sparkling from high level.

H2O is a fast and open-source machine learning tool for big data analysis. It was launched in Silicon Valley in 2011 by a company called, formerly called Oxdata. The company is leaded by a few top data scientists in the world, and also backed by a few mathematical professors at Standford University on the company’s scientific advisory board.

H2O uses in-memory compression and can handles billions of rows in-memory. It allows companies to use all of their data without sampling to get predication faster. It includes built-in advanced algorithms such as deep learning, boosting, and bagging ensembles. It allows organizations to build powerful domain-specific
predictive engines for recommendations, customer churn, propensity to buy, dynamic pricing, and fraud detection for insurance and credit card companies.

H2O has an interface to R, Scala, Python, and Java. Not only it can be run on Hadoop and Cloud environment (AWS, Google Cloud, and Azure), but also can run on Linux, Mac, and Windows. The following shows the H2O architecture.

For your own testing, you could install H2O on your laptop. For big dataset or accessing Spark Clusters are installed somewhere, use Sparking Water, which I will discuss in the later part of the blog. The installation of H2O on your laptop is super easy. Here are the steps:
1. Download the H2O Zip File
Goto H2O download page at, choose Latest Stable Release under H2O block.

2. Run H2O
Just unzip the file and then run the jar file. It will start a web server on your local machine.

cd h2o-
java -jar h2o.jar

3. Use H2O Flow UI
Input the link http://localhost:54321/flow/index.html in your browser and H2O Flow UI shows up as follows. You can do your data analysis work right now.

Sparking Water
Ok, let’s see what is Sparkling Water. In short, Sparking Water = H2O + Spark. Basically Sparking Water combines the fast machine learning algorithms of H2O with the popular in-memory platform – Spark to provide a fast and scalable solution for data analytics. Sparking Water supports Scala, R, or Python and can use H2O Flow UI to provide the machine learning platform for data scientists and application developers.

Sparking Water can run on the top of Spark in the following ways.

  • Local cluster
  • Standalone cluster
  • Spark cluster in a YARN environment

Sparking Water is designed as a Spark application. When it executes and you check from Spark UI, it is just a regular Spark Application. When it is launched, it first starts Spark Executors. Then H2O start services such as Key-Value store and memory manager inside executors. The following shows the relationship between Sparking Water, Spark and H2O.

To share data between Spark and H2O, Sparkling Water uses H2O’s H2OFrame. When converting an RDD/DataFrame to an H2O’s H2OFrame, it requires data duplication because it transfers data from RDD storage into H2OFrame. But data in H2OFrame is stored in compression format and does not need to be preserved in RDD.

A typical use case to use Sparking Water is to build Data Model. A model is constructed based on the estimation of metrics, testing data to give prediction that can be used in the rest of data pipeline.

The installation of Sparkling Water take a few more steps than H2O. As for now, Sparking Water is version 2.2.2. The detail installation steps are shown here at However, I don’t like the installation instruction for the following reasons:

  • It uses local Spark cluster as example. As far as I know, local Spark is not a typical way to use Spark. The environment variable setting to point to local Spark cluster is confusing. The easiest step should verify that spark-shell is working or not. If it works, skip the step to install spark cluster or set SPARK related environment variables.
  • It gives a simple instruction to run sparking-shell –conf “spark.executor.memory=1g”. It misses a lot of other parameters. Otherwise, it will give you tons of warning messages.

Ok, here are the steps I believe are the correct steps:
1. Verify Your Spark Environment
Of course, someone should have Spark cluster up and running before even considering Sparking Water installation. Installing a new Spark cluster without knowing whether the cluster is working or not will be a nightmare to trace Sparking Water and H2O water issue if the issue comes from Spark Cluster. Also make sure to install Spark2. Sparking Water and H2O has a very strict rule in terms of the version compatibility among Spark, Sparking Water, and H2O. Here is a brief overview of the matrix.

For detail information please check out

One easy way to verify spark cluster is good or not is to make sure you can access spark2-shell and run some simple example program without issues.

2. Download Sparking Water zip file

3. Unzip the file and run

cd sparkling-water-2.2.2

bin/sparkling-shell \
--master yarn \
--conf spark.executor.instances=12 \
--conf spark.executor.memory=10g \
--conf spark.driver.memory=8g \
--conf spark.scheduler.maxRegisteredResourcesWaitingTime=1000000 \
--conf \
--conf spark.dynamicAllocation.enabled=false \
--conf spark.sql.autoBroadcastJoinThreshold=-1 \
--conf spark.locality.wait=30000 \
--conf spark.scheduler.minRegisteredResourcesRatio=1

The above sample code allows you to create H2O cluster based on 12 executors with 10GB memory for each executor. You also have to add bunch of parameters like spark.sql.autoBroadcastJoinThreshold, spark.dynamicAllocation.enabled and many more. Setting spark.dynamicAllocation.enabled to false makes sense as you probably want to have H2O cluster to use resource in a predicable way and not suck in all cluster memory when processing large amount of data. Adding these parameters can help you to avoid many annoying warning message when sparking-shell starts.

If in kerberosed environment, make sure to run kinit before executing sparking-shell.

4. Create an H2O cluster inside Spark Cluster
Run the following code.

import org.apache.spark.h2o._
val h2oContext = H2OContext.getOrCreate(spark) 
import h2oContext._ 

The followings are the sample output from the run.

[install]$ cd sparkling-water-2.2.2
[sparkling-water-2.2.2]$ kinit wzhou
Password for
[sparkling-water-2.2.2]$ bin/sparkling-shell \
> --master yarn \
> --conf spark.executor.instances=12 \
> --conf spark.executor.memory=10g \
> --conf spark.driver.memory=8g \
> --conf spark.scheduler.maxRegisteredResourcesWaitingTime=1000000 \
> --conf \
> --conf spark.dynamicAllocation.enabled=false \
> --conf spark.sql.autoBroadcastJoinThreshold=-1 \
> --conf spark.locality.wait=30000 \
> --conf spark.scheduler.minRegisteredResourcesRatio=1

  Spark master (MASTER)     : yarn
  Spark home   (SPARK_HOME) :
  H2O build version         : (weierstrass)
  Spark build version       : 2.2.0
  Scala version             : 2.11

SLF4J: Class path contains multiple SLF4J bindings.
SLF4J: Found binding in [jar:file:/opt/anaconda2/lib/python2.7/site-packages/pyspark/jars/slf4j-log4j12-1.7.16.jar!/org/slf4j/impl/S                   taticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/opt/cloudera/parcels/CDH-5.10.1-1.cdh5.10.1.p0.10/jars/slf4j-log4j12-1.7.5.jar!/org/slf4j/impl/St                   aticLoggerBinder.class]
SLF4J: See for an explanation.
SLF4J: Actual binding is of type [org.slf4j.impl.Log4jLoggerFactory]
Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
17/11/05 07:46:05 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes w                   here applicable
17/11/05 07:46:05 WARN shortcircuit.DomainSocketFactory: The short-circuit local reads feature cannot be used because libhadoop cann                   ot be loaded.
17/11/05 07:46:06 WARN yarn.Client: Neither spark.yarn.jars nor spark.yarn.archive is set, falling back to uploading libraries under                    SPARK_HOME.
Spark context Web UI available at
Spark context available as 'sc' (master = yarn, app id = application_3608740912046_1362).
Spark session available as 'spark'.
Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /___/ .__/\_,_/_/ /_/\_\   version 2.2.0

Using Scala version 2.11.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_121)
Type in expressions to have them evaluated.
Type :help for more information.

scala> import org.apache.spark.h2o._
import org.apache.spark.h2o._

scala> val h2oContext = H2OContext.getOrCreate(spark)
h2oContext: org.apache.spark.h2o.H2OContext =

Sparkling Water Context:
 * H2O name: sparkling-water-wzhou_application_3608740912046_1362
 * cluster size: 12
 * list of used nodes:
  (executorId, host, port)

  Open H2O Flow in browser: (CMD + ...
scala> import h2oContext._
import h2oContext._

You can see H2O cluster is indeed inside a Spark application.

5. Access H2O cluster
You can access the H2O cluster in many ways. One is using the H2O Flow UI. Exact the same UI I show you before, like

Another way is to access it inside R Studio by using h2o.init(your_h2o_host, port). Use the above one as the example, here are the init command and other useful commands to check H2O cluster.

> library(sparklyr)
> library(h2o)
> library(dplyr)
> options(rsparkling.sparklingwater.version = "2.2.2")
> library(rsparkling)
> h2o.init(ip="", port=54325)
 Connection successful!

R is connected to the H2O cluster (in client mode): 
    H2O cluster uptime:         1 hours 7 minutes 
    H2O cluster version: 
    H2O cluster version age:    16 days  
    H2O cluster name:           sparkling-water-wzhou_application_3608740912046_1362 
    H2O cluster total nodes:    12 
    H2O cluster total memory:   93.30 GB 
    H2O cluster total cores:    384 
    H2O cluster allowed cores:  384 
    H2O cluster healthy:        TRUE 
    H2O Connection ip: 
    H2O Connection port:        54325 
    H2O Connection proxy:       NA 
    H2O Internal Security:      FALSE 
    H2O API Extensions:         Algos, AutoML, Core V3, Core V4 
    R Version:                  R version 3.3.2 (2016-10-31) 

> h2o.clusterIsUp()
[1] TRUE

> h2o.clusterStatus()
Cluster name: sparkling-water-wzhou_application_3608740912046_1362
Cluster size: 12 
Cluster is locked

                                             h2o healthy
1    TRUE
2    TRUE
3    TRUE
4    TRUE
5    TRUE
6    TRUE
7    TRUE
8    TRUE
9    TRUE
10    TRUE
11    TRUE
12    TRUE
      last_ping num_cpus sys_load mem_value_size   free_mem
1  1.509977e+12       32     1.21              0 8296753152
2  1.509977e+12       32     1.21              0 8615325696
3  1.509977e+12       32     0.63              0 8611375104
4  1.509977e+12       32     1.41              0 8290978816
5  1.509977e+12       32     0.33              0 8372216832
6  1.509977e+12       32     0.09              0 8164569088
7  1.509977e+12       32     0.23              0 8182391808
8  1.509977e+12       32     0.23              0 8579495936
9  1.509977e+12       32     0.15              0 8365195264
10 1.509977e+12       32     0.15              0 8131736576
11 1.509977e+12       32     0.63              0 8291118080
12 1.509977e+12       32     0.63              0 8243695616
     pojo_mem swap_mem    free_disk    max_disk   pid num_keys
1  1247908864        0 395950686208 4.91886e+11  4590        0
2   929336320        0 395950686208 4.91886e+11  4591        0
3   933286912        0 420010262528 4.91886e+11  3930        0
4  1253683200        0  4.25251e+11 4.91886e+11 16370        0
5  1172445184        0 425796304896 4.91886e+11 11998        0
6  1380092928        0 426025943040 4.91886e+11 20374        0
7  1362270208        0  4.21041e+11 4.91886e+11  4987        0
8   965166080        0  4.21041e+11 4.91886e+11  4988        0
9  1179466752        0 422286721024 4.91886e+11  6951        0
10 1412925440        0 422286721024 4.91886e+11  6952        0
11 1253543936        0 425969319936 4.91886e+11 24232        0
12 1300966400        0 425969319936 4.91886e+11 24233        0
   tcps_active open_fds rpcs_active
1            0      453           0
2            0      452           0
3            0      453           0
4            0      453           0
5            0      452           0
6            0      453           0
7            0      452           0
8            0      452           0
9            0      452           0
10           0      453           0
11           0      453           0
12           0      452           0

> h2o.networkTest()
Network Test: Launched from
1                         all - collective bcast/reduce
2  remote
3  remote
4  remote
5  remote
6  remote
7  remote
8  remote
9  remote
10 remote
11 remote
12 remote
13 remote
                   1_bytes               1024_bytes
1   38.212 msec,  628  B/S     6.790 msec, 3.5 MB/S
2    4.929 msec,  405  B/S       752 usec, 2.6 MB/S
3    7.089 msec,  282  B/S       710 usec, 2.7 MB/S
4    5.687 msec,  351  B/S       634 usec, 3.1 MB/S
5    6.623 msec,  301  B/S       784 usec, 2.5 MB/S
6    6.277 msec,  318  B/S   2.680 msec, 746.0 KB/S
7    6.469 msec,  309  B/S       840 usec, 2.3 MB/S
8    6.595 msec,  303  B/S       801 usec, 2.4 MB/S
9    5.155 msec,  387  B/S       793 usec, 2.5 MB/S
10   5.204 msec,  384  B/S       703 usec, 2.8 MB/S
11   5.511 msec,  362  B/S       782 usec, 2.5 MB/S
12   6.784 msec,  294  B/S       927 usec, 2.1 MB/S
13   6.001 msec,  333  B/S       711 usec, 2.7 MB/S
1   23.800 msec, 1008.4 MB/S
2     6.997 msec, 285.8 MB/S
3     5.576 msec, 358.6 MB/S
4     4.056 msec, 493.0 MB/S
5     5.066 msec, 394.8 MB/S
6     5.272 msec, 379.3 MB/S
7     5.176 msec, 386.3 MB/S
8     6.831 msec, 292.8 MB/S
9     5.772 msec, 346.4 MB/S
10    5.125 msec, 390.2 MB/S
11    5.274 msec, 379.2 MB/S
12    5.065 msec, 394.9 MB/S
13    4.960 msec, 403.2 MB/S

Create Cloudera Hadoop Cluster Using Cloudera Director on Google Cloud

I have a blog discussing how to install Cloudera Hadoop Cluster several years ago. It basically took about at least half day to complete the installation in my VM cluster. In my last post, I discussed an approach to deploy Hadoop cluster using DataProc on Google Cloud Platform. It literally took less than two minutes to create a Hadoop Cluster. Although it is a good to have a cluster launched in a very short time, it does not have the nice UI like Cloudera Manager as the Hadoop distribution used by Dataproc is not CDH. I could repeat my blogs to build a Hadoop Cluster using VM instances on Google Cloud Platform. But it will take some time and involve a lot of work. Actually there is another way to create Hadoop cluster on the cloud. Cloudera has a product, called Cloudera Director. It currently supports not only Google Cloud, but also AWS and Azure as well. It is designed to deploy CDH cluster faster and easier to scale the cluster on the cloud. Another important feature is that Cloud Director allows you to move your deployment scripts or steps easily from one cloud provider to another provider and you don’t have to be locked in one cloud vendor. In this blog, I will show you the way to create a CDH cluster using Cloudera Director.

The first step is to start my Cloudera Director instance. In my case, I have already installed Cloudera Director based on the instruction from Cloudera. It is pretty straight forward process and I am not going to repeat it here. The Cloudera Director instance is where you can launch your CDH cluster deployment.

Both Cloudera Director and Cloudera Manager UI are browser-based and you have to setup secure connection between your local machine and VM instances on the cloud. To achieve this, you need to configure SOCKS proxy on your local machine that is used to connect to the Cloudera Director VM. It provides a secure way to connect to your VM on the cloud and can use VM’s internal IP and hostname in the web browser. Google has a nice note about the steps, Securely Connecting to VM Instances. Following this note will help you to setup SOCKS proxy.

Ok, here are the steps.
Logon to Cloudera Director
Open a terminal session locally, and run the following code:

gcloud compute ssh cdh-director-1 \
    --project cdh-director-173715 \
    --zone us-central1-c \
    --ssh-flag="-D" \
    --ssh-flag="1080" \

cdh-director-1 is the name of my Cloudera Director instance on Google cloud and cdh-director-173715 is my Google Cloud project id. After executing the above command, it looks hang and never complete. This is CORRECT behavior. Do not kill or exit this session. Open a browser and type in the internal IP of Cloudera Director instance with port number 7189. For my cdh-director-1 instance, the internal IP is

After input the URL for Cloudera Director. The login screen shows up. Login as admin user.

After login, the initial setup wizard shows up. Click Let’s get started.

In the Add Environment screen, input the information as follows. The Client ID JSON Key is the file you can create during the initial setup of you Google project with SSH key stuff.

In the next Add Cloudera Manager screen, I usually create the Instance Template first. Click the drop down of Instance Template, then select Create a new instance template. I need at least three template, one for Cloudera Manager, one for Master nodes, and one for Worker nodes. In my case here, I did not create a template for Edge nodes. To save resource on my Google cloud environment, I did not create the template for Edge node. Here are the configuration for all three templates.

Cloudera Manager Template

Master Node Template

Worker Node Template

Input the following for Cloudera Manager. For my test, I use Embedded Database. If it is used for production, you need to setup external database first and register the external database here.

After click Continue, Add Cluster screen shows up. There is a gateway instance group and I removed it by clicking Delete Group because I don’t have edge node here. Input the corresponding template and number of instances for masters and workders.

After click Continue, the deployment starts.

After about 20 minutes, it completes. Click Continue.

Review Cluster
The nice Cloudera Director dashboard shows up.

You can also login to Cloudera Manager from the link on Cloudera Director.

Nice and easy. Excellent product from Cloudera. For more information about deploying CDH cluster on Google Cloud, you can also check out Cloudera’s document, Getting Started on Google Cloud Platform.

Resolve Sparklyr not Respond Issue on Port 8880

Recently I was approached by one of my clients to help them to investigate a weird Sparklyr issue. sparklyr is an interface between R and Spark introduced by RStudio about a years ago. The following is the the sparklyr architecture.

When trying to do sc <- spark_connect in RStudio, we got two errors as follows:

  • Failed while connecting to sparklyr to port (8880) for sessionid (3859): Gateway in port (8880) did not respond.
  • Exception in thread “main” java.lang.NoClassDefFoundError: org/apache/hadoop/fs/FSDataInputStream
  • Here is the detail message.

    > library(sparklyr)
    > library(dplyr) 
    > sc <- spark_connect(master = "yarn-client", config=spark_config(), version="1.6.0", spark_home = '/opt/cloudera/parcels/CDH/lib/spark/')
    Error in force(code) :
    Failed while connecting to sparklyr to port (8880) for sessionid (3859): Gateway in port (8880) did not respond.
    Path: /opt/cloudera/parcels/CDH-5.10.1-1.cdh5.10.1.p0.10/lib/spark/bin/spark-submit
    Parameters: --class, sparklyr.Shell, --jars, '/usr/lib64/R/library/sparklyr/java/spark-csv_2.11-1.3.0.jar','/usr/lib64/R/library/sparklyr/java/commons-csv-1.1.jar','/usr/lib64/R/library/sparklyr/java/univocity-parsers-1.5.1.jar', '/usr/lib64/R/library/sparklyr/java/sparklyr-1.6-2.10.jar', 8880, 3859
    Log: /tmp/RtmzpSIMln/file9e23246605df7_spark.log
    Exception in thread "main" java.lang.NoClassDefFoundError: org/apache/hadoop/fs/FSDataInputStream
                   at org.apache.spark.deploy.SparkSubmitArguments.handle(SparkSubmitArguments.scala:394)
                   at org.apache.spark.launcher.SparkSubmitOptionParser.parse(
                   at org.apache.spark.deploy.SparkSubmitArguments.<init>(SparkSubmitArguments.scala:97)
                   at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:114)
                   at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)
    Caused by: java.lang.ClassNotFoundException: org.apache.hadoop.fs.FSDataInputStream
                   at java.lang.ClassLoader.loadClass(
                   at sun.misc.Launcher$AppClassLoader.loadClass(
                   at java.lang.ClassLoader.loadClass(
                   ... 5 more

    Did some research and found many people having the similar issue. Ok, try their recommendations one by one as follows.

  • Set SPARK_HOME environment
  • Try run Sys.setEnv(SPARK_HOME = “/opt/cloudera/parcels/CDH/lib/spark/”). No, not working.

  • Install latest version sparklyr
  • My client installed sparklyr less than one month ago. I don’t see why this option makes sense. Don’t even pursue this path.

  • Check Java Installation
  • The R on the same server uses the same version of Java without any issue. I don’t see why Java installation become a major concern here. Ignore this one.

  • No Hadoop Installation
  • Someone said just Spark installation is not enough, not to have Hadoop Installation as well. Clearly it does not fit our situation. The server is an edge node and has hadoop installation.

  • Do not have a valid kerberos ticket
  • Running system2(‘klist’) does show no kerberos ticket. Ok, I then open up a shell within RStudio Server by clicking tools -> shell, then issuing the kinit command.
    Rerun system2(‘klist’) shows I have a valid kerberos ticket. Try again. still not working.
    Note: even it is not working, this step is necessary for further action when the issue is fixed. So still need to run this one no matter what the result is.

  • Create a different configure and pass to spark_connect
  • Someone recommended to create a new configure and pass it in. It looks like a good idea. Unfortunately, just doesn’t work.

    wzconfig <- spark_config()
    wzconfig$`` <- "client"
    wzconfig$spark.driver.cores <- 1
    wzconfig$spark.executor.cores <- 2
    wzconfig$spark.executor.memory <- "4G"
    sc <- spark_connect(master = "yarn-client", config=wzconfig, version="1.6.0", spark_home = '/opt/cloudera/parcels/CDH/lib/spark/')

    Actually this recommendation is missing another key parameter. By default the total number of executors launched is 2. I would usually bump up this number a little to get a better performance. You can use the following way to set up the
    total number of executors.

    wzconfig$spark.executor.instances <- 3

    Although this approach looks promising, still not working. But this approach is definitely a way to use for other purpose to better control the Spark resource usage.

  • Add remote address
  • Someone mentioned to set remote address. I thought this could another potential option as I resolved issues in Spark related to local IP issue in the past. So I add the following code in the configuration from the previous example, note parameter sparklyr.gateway.address is the hostname of active Resource Manager.

    wzconfig$sparklyr.gateway.remote <- TRUE
    wzconfig$sparklyr.gateway.address <- "" 

    Not working for this case.

  • Change deployment mode to yarn-cluster
  • This is probably the most unrealistic one. If connect as with master = “yarn-cluster”, the spark driver will be somewhere inside the Spark cluster. For our current case, I don’t believe this is the right solution. Don’t even try it.

  • Run Spark example
  • Someone recommended to run a spark-submit to verify SparkPi can be run from the environment. This looks reasonable. The good thing I figured out the issue before executing this one. But this definitely a valid and good test to verify spark-submit.

    /opt/cloudera/parcels/SPARK2/lib/spark2/bin/spark-submit --class org.apache.spark.examples.SparkPi --deploy-mode client --master yarn /opt/cloudera/parcels/SPARK2/lib/spark2/examples/jars/spark-examples_2.11-2.1.0.jar 10
  • HA for yarn-cluster
  • There is an interesting post Add support for `yarn-cluster` with high availability #905 discussing about the issue might relate to multiple resource managers. We use HA and this post is an interesting one. But might not fit into our case because I feel we have not reached to the HA part yet with Class Not Found message.

  • Need to set JAVA_HOME
  • Verified it and we have it. So this is not the issue.

  • My Solution
  • After reviewing or trying out some of above solutions, I like to go back my way of thinking. I must say I am not an expert in R or RStudio with very limited knowledge about how it works. But I did have extensive background in Spark tuning and trouble shooting.

    I know the error message Gateway in port (8880) did not respond is always the first message shows up and looks like the cause of the issue. But I thought differently. I believe the 2nd error NoClassDefFoundError: org/apache/hadoop/fs/FSDataInputStream looks more suspicious than the first one. Early this year I helped one of another clients on a weird Spark job issue, which is in the end, was caused by the incorrect path. It seems to me the path might not be right and cause Spark issue, then caused the first error of port not respond.

    With this idea in mind, I focused more the path verification. Run the command Sys.getenv() to get the environment as follows.

    > Sys.getenv()
    DISPLAY                 :0
    EDITOR                  vi
    GIT_ASKPASS             rpostback-askpass
    HADOOP_CONF_DIR         /etc/hadoop/conf.cloudera.hdfs
    HADOOP_HOME             /opt/cloudera/parcels/CDH
    HOME                    /home/wzhou
    JAVA_HOME               /usr/java/jdk1.8.0_144/jre
    LANG                    en_US.UTF-8
    LD_LIBRARY_PATH         /usr/lib64/R/lib::/lib:/usr/java/jdk1.8.0_92/jre/lib/amd64/server
    LN_S                    ln -s
    LOGNAME                 wzhou
    MAKE                    make
    PAGER                   /usr/bin/less
    PATH                    /usr/local/sbin:/usr/local/bin:/usr/bin:/usr/sbin:/sbin:/bin
    R_BROWSER               /usr/bin/xdg-open
    R_BZIPCMD               /usr/bin/bzip2
    R_DOC_DIR               /usr/share/doc/R-3.3.2
    R_GZIPCMD               /usr/bin/gzip
    R_HOME                  /usr/lib64/R
    R_INCLUDE_DIR           /usr/include/R
    R_LIBS_SITE             /usr/local/lib/R/site-library:/usr/local/lib/R/library:/usr/lib64/R/library:/usr/share/R/library
    R_LIBS_USER             ~/R/x86_64-redhat-linux-gnu-library/3.3
    R_PAPERSIZE             a4
    R_PDFVIEWER             /usr/bin/xdg-open
    R_PLATFORM              x86_64-redhat-linux-gnu
    R_PRINTCMD              lpr
    R_RD4PDF                times,hyper
    R_SESSION_TMPDIR        /tmp/RtmpZf9YMN
    R_SHARE_DIR             /usr/share/R
    R_SYSTEM_ABI            linux,gcc,gxx,gfortran,?
    R_TEXI2DVICMD           /usr/bin/texi2dvi
    R_UNZIPCMD              /usr/bin/unzip
    RS_RPOSTBACK_PATH       /usr/lib/rstudio-server/bin/rpostback
    RSTUDIO                 1
    RSTUDIO_PANDOC          /usr/lib/rstudio-server/bin/pandoc
    RSTUDIO_WINUTILS        bin/winutils
    SED                     /bin/sed
    SPARK_HOME              /opt/cloudera/parcels/SPARK2/lib/spark2
    SSH_ASKPASS             rpostback-askpass
    TAR                     /bin/gtar
    USER                    wzhou
    YARN_CONF_DIR           /etc/hadoop/conf.cloudera.yarn

    Ahhh, I noticed the environment missed SPARK_DIST_CLASSPATH environment variable. Then I set it using the command below just before sc <- spark_connect.

    Sys.setenv(SPARK_DIST_CLASSPATH = '/etc/hadoop/con:/opt/cloudera/parcels/CDH/lib/hadoop/libexec/../../hadoop/lib/*:/opt/cloudera/parcels/CDH/lib/hadoop/libexec/../../hadoop/.//*:/opt/cloudera/parcels/CDH/lib/hadoop/libexec/../../hadoop-hdfs/./:/opt/cloudera/parcels/CDH/lib/hadoop/libexec/../../hadoop-hdfs/lib/*:/opt/cloudera/parcels/CDH/lib/hadoop/libexec/../../hadoop-hdfs/.//*:/opt/cloudera/parcels/CDH/lib/hadoop/libexec/../../hadoop-yarn/lib/*:/opt/cloudera/parcels/CDH/lib/hadoop/libexec/../../hadoop-yarn/.//*:/opt/cloudera/parcels/CDH/lib/hadoop-mapreduce/lib/*:/opt/cloudera/parcels/CDH/lib/hadoop-mapreduce/.//*')

    Ok, try it again. Fantastic, it works!

    Ok, here is the real cause of the issue. It’s unnecessary to specify java path for sparklyr as it does not require a java path. However, it does have dependency on spark-submit. When spark-submit is executed, it can read java path and then submit the jar files to Spark accordingly. The cause of the issue if SPARK_DIST_CLASSPATH is not set, spark-submit is not working and Spark executors can not be launched.

    Other Note
    The following are some of useful commands:

    spark_home_dir() or spark_home
    config <- spark_config()


    Also there are a few useful articles about sparklyr and Rstudio:
    RStudio’s R Interface to Spark on Amazon EMR
    How to Install RStudio Server on CentOS 7
    Using R with Apache Spark
    sparklyr: a test drive on YARN
    Analyzing a billion NYC taxi trips in Spark

    Create Hadoop Cluster on Google Cloud Platform

    There are many ways to create Hadoop clusters and I am going to show a few ways on Google Cloud Platform (GCP). The first approach is the standard way to build a Hadoop cluster, no matter whether you do it on cloud or on-premise. Basically create a group of VM instances and manually install Hadoop cluster on these VM instances. Many people have blogs or articles about this approach and I am not going to repeat the steps here.

    In this blog, I am going to discuss the approach using Google Cloud Dataproc and you can actually have a Hadoop cluster up and running
    within 2 minutes. Google Cloud Dataproc is a fully-managed cloud service for running Apache Hadoop cluster in a simple and fast way. The followings show the steps to create a Hadoop Cluster and submit a spark job to the cluster.

    Create a Hadoop Cluster
    Click Dataproc -> Clusters

    Then click Enable API

    Cloud Dataproc screen shows up. Click Create cluster

    Input the following parameters:
    Name : cluster-test1
    Region : Choose use-central1
    Zone : Choose us-central1-c

    1. Master Node
    Machine Type : The default is n1-standard-4, but I choose n1-standard-1 just for simple testing purpose.
    Cluster Mode : There are 3 modes here. Single Mode (1 master, 0 worker), Standard Mode (1 master, N worker), and High Mode (3 masters, N workers). Choose Standard Mode.
    Primary disk size : For my testing, 10GB 1s enough.

    2. Worker Nodes
    Similar configuration like Worker node. I use 3 worker nodes and disk size is 15 GB. You might notice that there is option to use local SSD storage. You can attach up to 8 local SSD devices to the VM instance. Each disk is 375 GB in size and you can not specify 10GB disk size here. The local SSDs are physically attached to the host server and offer higher performance and lower latency storage than Google’s persistent disk storage. The local SSDs is used for temporary data like shuffling data in MapReduce. The data on the local SSD storage is not persistent. For more information, please visit

    Another thing to mention is that Dataproc uses Cloud Storage bucket instead of HDFS for the Hadoop cluster. Although the hadoop command is still working and you won’t feel anything different, the underline storage is different. In my opinion, it is actually better because Google Cloud Storage bucket definitely has much better reliability and scalability than HDFS.

    Click Create when everything is done. After a few minutes, the cluster is created.

    Click cluster-test1 and it should show the cluster information.

    If click VM Instances tab, we can see there is one master and 3 worker instances.

    Click Configuration tab. It shows all configuration information.

    Submit a Spark Job
    Click Cloud Dataproc -> Jobs.

    Once Submit Job screen shows up, input the following information, then click Submit.

    After the job completes, you should see the followings:

    To verify the result, I need to ssh to the master node to find out which ports are listening for connections. Click the drop down on the right of SSH of master node, then click Open in browser window.

    Then run the netstat command.

    cluster-test1-m:~$ netstat -a |grep LISTEN |grep tcp
    tcp        0      0 *:10033                 *:*                     LISTEN     
    tcp        0      0 *:10002                 *:*                     LISTEN     
    tcp        0      0 cluster-test1-m.c.:8020 *:*                     LISTEN     
    tcp        0      0 *:33044                 *:*                     LISTEN     
    tcp        0      0 *:ssh                   *:*                     LISTEN     
    tcp        0      0 *:52888                 *:*                     LISTEN     
    tcp        0      0 *:58266                 *:*                     LISTEN     
    tcp        0      0 *:35738                 *:*                     LISTEN     
    tcp        0      0 *:9083                  *:*                     LISTEN     
    tcp        0      0 *:34238                 *:*                     LISTEN     
    tcp        0      0 *:nfs                   *:*                     LISTEN     
    tcp        0      0 cluster-test1-m.c:10020 *:*                     LISTEN     
    tcp        0      0 localhost:mysql         *:*                     LISTEN     
    tcp        0      0 *:9868                  *:*                     LISTEN     
    tcp        0      0 *:9870                  *:*                     LISTEN     
    tcp        0      0 *:sunrpc                *:*                     LISTEN     
    tcp        0      0 *:webmin                *:*                     LISTEN     
    tcp        0      0 cluster-test1-m.c:19888 *:*                     LISTEN     
    tcp6       0      0 [::]:10001              [::]:*                  LISTEN     
    tcp6       0      0 [::]:44884              [::]:*                  LISTEN     
    tcp6       0      0 [::]:50965              [::]:*                  LISTEN     
    tcp6       0      0 [::]:ssh                [::]:*                  LISTEN     
    tcp6       0      0 cluster-test1-m:omniorb [::]:*                  LISTEN     
    tcp6       0      0 [::]:46745              [::]:*                  LISTEN     
    tcp6       0      0 cluster-test1-m.c.:8030 [::]:*                  LISTEN     
    tcp6       0      0 cluster-test1-m.c.:8031 [::]:*                  LISTEN     
    tcp6       0      0 [::]:18080              [::]:*                  LISTEN     
    tcp6       0      0 cluster-test1-m.c.:8032 [::]:*                  LISTEN     
    tcp6       0      0 cluster-test1-m.c.:8033 [::]:*                  LISTEN     
    tcp6       0      0 [::]:nfs                [::]:*                  LISTEN     
    tcp6       0      0 [::]:33615              [::]:*                  LISTEN     
    tcp6       0      0 [::]:56911              [::]:*                  LISTEN     
    tcp6       0      0 [::]:sunrpc             [::]:*                  LISTEN  

    Check out directories.

    cluster-test1-m:~$ hdfs dfs -ls /
    17/09/12 12:12:24 INFO gcs.GoogleHadoopFileSystemBase: GHFS version: 1.6.1-hadoop2
    Found 2 items
    drwxrwxrwt   - mapred hadoop          0 2017-09-12 11:56 /tmp
    drwxrwxrwt   - hdfs   hadoop          0 2017-09-12 11:55 /user

    There are a few UI screens available to check out the Hadoop cluster and job status.
    HDFS NameNode (port 9870)

    YARN Resource Manager (port 8088)

    Spark Job History (port 18080)

    Dataproc approach is an easy deployment tool to create a Hadoop cluster. Although it is powerful, I miss the nice UI like Cloudera Manager. To install Cloudera CDH cluster, I need to use a different approach and I am going to discuss it in the future blog.