Tag Archives: Spark

Copy Data with Hive and Spark / Copiar Datos con Hive y Spark

Posted on by
Reply

These are two examples of how to copy data from one S3 location to other S3 location. Same operation can be done from S3 to HDFS and vice-versa.

I’m considering that you are able to launch the Hive client or spark-shell client.

Hive:

Using Mapreduce engine or Tez engine:

set hive.execution.engine=mr;

or

set hive.execution.engine=tez; 
CREATE EXTERNAL TABLE source_table(a_col string, b_col string, c_col string)
ROW FORMAT DELIMITED FIELDS TERMINATED BY ','
LOCATION 's3://mybucket/hive/csv/';

CREATE TABLE destination_table(a_col string, b_col string, c_col string) LOCATION 's3://mybucket/output-hive/csv_1/';

INSERT OVERWRITE TABLE destination_table SELECT * FROM source_table;

Spark:

sc.textFile("s3://aws-publicdatasets/common-crawl/crawl-data/CC-MAIN-2013-48/segments/1387346051826/warc/").saveAsTextFile("s3://mybucket/spark/bigfiles")

 

If you want to copy data to HDFS, you can also explore s3-dist-cp:

s3DistCP:

s3-dist-cp --src s3://mybucket/hive/csv/ --dest=hdfs:///output-hive/csv_10/

 

get the driver’s IP in spark yarn-cluster mode

Posted on by
Reply

In cluster mode, the Spark driver runs inside an application master process which is managed by YARN on the cluster, and the client can go away after initiating the application. In client mode, the driver runs in the client process, and the application master is only used for requesting resources from YARN.

Sometimes we will have a bunch of logs for a terminated cluster and we need to find out which node was the driver in cluster mode.

Searching for “driverUrl” on the application/container logs, we will find it

find . -iname "*.gz" | xargs zgrep "driverUrl"
./container_1459071485818_0006_02_000001/stderr.gz:15/03/28 05:10:47 INFO YarnAllocator: Launching ExecutorRunnable. driverUrl: spark://CoarseGrainedScheduler@172.31.16.15:47452,  executorHostname: ip-172-31-16-13.ec2.internal
...
./container_1459071485818_0006_02_000001/stderr.gz:15/03/28 05:10:47 INFO YarnAllocator: Launching ExecutorRunnable. driverUrl: spark://CoarseGrainedScheduler@172.31.16.15:47452,  executorHostname: ip-172-31-16-14.ec2.internal

On this case the driver was running on 172.31.16.15.

Consider boosting spark.yarn.executor.memoryOverhead

Posted on by
Reply

This is a very specific error related to the Spark Executor and the YARN container coexistence.

You will typically see errors like this one on the application container logs:

15/03/12 18:53:46 WARN YarnAllocator: Container killed by YARN for exceeding memory limits. 9.3 GB of 9.3 GB physical memory used. Consider boosting spark.yarn.executor.memoryOverhead.
15/03/12 18:53:46 ERROR YarnClusterScheduler: Lost executor 21 on ip-xxx-xx-xx-xx: Container killed by YARN for exceeding memory limits. 9.3 GB of 9.3 GB physical memory used. Consider boosting spark.yarn.executor.memoryOverhead.

To overcome this, you need to keep in mind how Yarn container and the executor are set in memory:

spark-tuning-yarn-memory

Memory used by Spark Executor is exceeding the predefined limits (often caused by a few spikes) and that is causing YARN to kill the container with the previously mentioned message error.

By default ‘spark.yarn.executor.memoryOverhead’ parameter is set to 384 MB. This value could be low depending on your application and the data load.

Suggested value for this parameter is ‘executorMemory * 0.10’.

We can increase the value for ‘spark.yarn.executor.memoryOverhead’ to 1GB on spark-submit bu adding this to the command line:

–conf spark.yarn.executor.memoryOverhead=1024

For reference, this fix was added on Jira 1930:

+  <td><code>spark.yarn.executor.memoryOverhead</code></td>

 

Running Spark with oozie

Posted on by
Reply

Oozie 4.2 now supports spark-action.

Example job.properties file (configuration tested on EMR 4.2.0):

nameNode=hdfs://172.31.25.17:8020 
jobTracker=172.31.25.17:8032 
master=local[*] 
queueName=default 
examplesRoot=examples 
oozie.use.system.libpath=true 
oozie.wf.application.path=${nameNode}/user/${user.name}/${examplesRoot}/apps/spark

(Use the master node internal IP instead of localhost in the nameNode and jobTracker)

Validate oozie workflow xml file:

oozie validate workflow.xml

Example workflow.xml file:

<workflow-app xmlns='uri:oozie:workflow:0.5' name='SparkFileCopy'>
 <start to='spark-node' />
<action name='spark-node'>
 <spark xmlns="uri:oozie:spark-action:0.1">
 <job-tracker></job-tracker>
 <name-node></name-node>
 <prepare>
 <delete path="/user/${wf:user()}//output-data/spark"/>
 </prepare>
 <master></master>
 <name>Spark-FileCopy</name>
 <class>org.apache.oozie.example.SparkFileCopy</class>
 <jar>/user/${wf:user()}//apps/spark/lib/oozie-examples.jar</jar>
 <arg>/user/${wf:user()}//input-data/text/data.txt</arg>
 <arg>/user/${wf:user()}//output-data/spark</arg>
 </spark>
 <ok to="end" />
 <error to="fail" />
 </action>
<kill name="fail">
 <message>Workflow failed, error
 message[${wf:errorMessage(wf:lastErrorNode())}]
 </message>
 </kill>
 <end name='end' />
 </workflow-app>

Create the defined structure in HDFS and copy the proper files:

hadoop fs -ls /user/hadoop/examples/apps/spark/
Found 3 items
drwxr-xr-x - hadoop hadoop 0 2015-12-18 08:13 /user/hadoop/examples/apps/spark/lib
-rw-r--r-- 1 hadoop hadoop 1920 2015-12-18 08:08 /user/hadoop/examples/apps/spark/workflow.xml

hadoop fs -put workflow.xml /user/hadoop/examples/apps/spark/

hadoop fs -put /usr/share/doc/oozie-4.2.0/examples/apps/spark/lib/oozie-examples.jar /user/hadoop/examples/apps/spark/lib

hadoop fs -mkdir -p /user/hadoop/examples/input-data/text

hadoop fs -mkdir -p /user/hadoop/examples/output-data/spark

hadoop fs -put /usr/share/doc/oozie-4.2.0/examples/input-data/text/data.txt /user/hadoop/examples/input-data/text/

 

Run your oozie Job:

oozie job --oozie http://localhost:11000/oozie -config ./job.properties -run

Check oozie job:

oozie job -info 0000004-151203092421374-oozie-oozi-W

Check available sharelib:

$ oozie admin -shareliblist -oozie http://localhost:11000/oozie
[Available ShareLib] 
oozie 
hive 
distcp 
hcatalog 
sqoop 
mapreduce-streaming 
spark 
hive2 
pig

 

 

References:

https://oozie.apache.org/docs/4.2.0/DG_SparkActionExtension.html

 

Apache Spark – How it Works – Performance Notes

Posted on by
1

Apache Spark, is an open source cluster computing framework originally developed at University of California, Berkeley but was later donated to the Apache Software Foundation where it remains today. In contrast to Hadoop’s two-stage disk-based MapReduce paradigm, Spark’s multi-stage in-memory primitives provides performance up to 100 faster for certain applications.

RDD’s:

Spark has a driver program where the application logic execution is started, with multiple workers which processing data in parallel.
The data is typically collocated with the worker and partitioned across the same set of machines within the cluster.  During the execution, the driver program will pass the code/closure into the worker machine where processing of corresponding partition of data will be conducted.
The data will undergoing different steps of transformation while staying in the same partition as much as possible (to avoid data shuffling across machines).  At the end of the execution, actions will be executed at the worker and result will be returned to the driver program.

spark_driver

Spark revolves around the concept of a resilient distributed dataset (RDD), which is a fault-tolerant collection of elements that can be operated on in parallel.

spark_rdd

There are currently two types of RDDs:

  • parallelized collections, which take an existing Scala collection and run functions on it in parallel.
  • Hadoop datasets, which run functions on each record of a file in Hadoop distributed file system or any other storage system supported by Hadoop.

Both types of RDDs can be operated on through the same methods. Each application has a driver process which coordinates its execution.

Running Applications:

Spark applications are similar to MapReduce “jobs”. Each application is a self-contained computation which runs some user-supplied code to compute a result. As with MapReduce jobs, Spark applications can make use of the resources of multiple nodes.

This process can run in the foreground (client mode) or in the background (cluster mode). Client mode is a little simpler, but cluster mode allows you to easily log out after starting a Spark application without terminating the application.

Spark starts executors to perform computations. There may be many executors, distributed across the cluster, depending on the size of the job. After loading some of the executors, Spark attempts to match tasks to executors.

Spark can run in two modes:

  • Standalone mode: Spark uses a Master daemon which coordinates the efforts of the Workers, which run the executors. Standalone mode is the default, but it cannot be used on secure clusters.
  • YARN mode: The YARN ResourceManager performs the functions of the Spark Master. The functions of the Workers are performed by the YARN NodeManager daemons, which run the executors. YARN mode is slightly more complex to set up, but it supports security, and provides better integration with YARN’s cluster-wide resource management policies.

Spark Execution Parameters

–num-executors: The –num-executors command-line flag or spark.executor.instances configuration property control the number of executors requested

–executor-cores: This property controls the number of concurrent tasks an executor can run. –executor-cores 5 means that each executor can run a maximum of five tasks at the same time.

Every Spark executor in an application has the same fixed number of cores and same fixed heap size.

The number of cores can be specified with the –executor-cores flag when invoking spark-submit, spark-shell, and pyspark from the command line.

The heap size can be controlled with the –executor-cores flag or the spark.executor.memory property.

–executor-memory: This property controls the executor heap size, but JVMs can also use some memory off heap, for example for interned Strings and direct byte buffers.

The value of the spark.yarn.executor.memoryOverhead property is added to the executor memory to determine the full memory request to YARN for each executor.

It defaults to max(384, .07 * spark.executor.memory).

Application Master:

Is a non-executor container with the special capability of requesting containers from YARN, takes up resources of its own that must be budgeted in. In yarn-client mode, it defaults to a 1024MB and one vcore. In yarn-cluster mode, the application master runs the driver, so it’s often useful to bolster its resources with the –driver-memory and –driver-cores properties.

Performance – Determining the number of executors:

It’s important to think about how the resources requested by Spark will fit into what YARN has available:
– yarn.nodemanager.resource.memory-mb controls the maximum sum of memory used by the containers on each node.
– yarn.nodemanager.resource.cpu-vcores controls the maximum sum of cores used by the containers on each node.

spark-tuning-yarn-memory

As an example, in a cluster with six nodes running NodeManagers, each equipped with 16 cores and 64GB of memory:
The NodeManager capacities, yarn.nodemanager.resource.memory-mb and yarn.nodemanager.resource.cpu-vcores, should probably be set to: 63 * 1024 = 64512 (megabytes) and 15 respectively

We must avoid allocating 100% of the resources to YARN containers because the node needs some resources to run the OS and Hadoop daemons.

In this case, we leave a gigabyte and a core for these system processes.

An executors configuration approach could be:
 –num-executors 17 –executor-cores 5 –executor-memory 19G.

This config results in 3 executors per node, except for the one with the AM, which will have 2 executors.
(executor-memory was derived as (63/3 executors per node) = 21.  21 * 0.07 = 1.47.  21 – 1.47 ~ 19)

Conclusions:

  • Depending on the application size (memory), using small executors (several executors per node) will perform better.
  • If the executors are too tiny (with a single core and just enough memory needed to run a single task, for example) throws away the benefits that come from running multiple tasks in a single JVM.
  • The application master, which is a non-executor container with the special capability of requesting containers from YARN, takes up resources of its own that must be budgeted in. In yarn-client mode, it defaults to a 1024MB and one vcore. In yarn-cluster mode, the application master runs the driver, so it’s often useful to bolster its resources with the --driver-memory and --driver-cores properties.

spark-yarn-f22

  • In yarn-cluster mode, the application master runs the driver, so it’s useful to bolster its resources with the –driver-memory and –driver-cores properties.
  • Running executors with too much memory often results in excessive garbage collection delays.

References:

http://horicky.blogspot.ie/2013/12/spark-low-latency-massively-parallel.html
http://blog.cloudera.com/blog/2015/03/how-to-tune-your-apache-spark-jobs-part-2/
https://spark.apache.org/docs/latest/tuning.html