For various reasons I'm running some jobs on EMR with AMI 2.4.11/Hadoop 1.0.3. I'm trying to run a cleanup of HDFS after my jobs by adding an additional EMR step. Using boto:
step = JarStep(
'HDFS cleanup',
'command-runner.jar',
action_on_failure='CONTINUE',
step_args=['hadoop', 'dfs', '-rmr', '-skipTrash', 'hdfs:/tmp'])
emr_conn.add_jobflow_steps(cluster_id, [step])
However it regularly fails with nothing in stderr in the EMR console.
Why I am confused is if I ssh into the master node and run the command:
hadoop dfs -rmr -skipTrash hdfs:/tmp
It succeeds with a 0 and a message that it successfully deleted everything. All the normal hadoop commands seem to work as documented. Does anyone know if there's an obvious reason for this? Issue with the Amazon distribution? Undocumented behavior in certain commands?
Note:
I have other jobs running in Hadoop 2 and the documented:
hdfs dfs -rm -r -skipTrash hdfs:/tmp
works as one would expect both as a step and as a command.
My solution generally was to upgrade everything to Hadoop2, in which case this works:
JarStep(
'%s: HDFS cleanup' % self.job_name,
'command-runner.jar',
action_on_failure='CONTINUE',
step_args=['hdfs', 'dfs', '-rm', '-r', '-skipTrash', path]
)
This was the best I could get with Hadoop1 that worked pretty well.
JarStep(
'%s: HDFS cleanup' % self.job_name,
'command-runner.jar',
action_on_failure='CONTINUE',
step_args=['hadoop', 'fs', '-rmr', '-skipTrash',
'hdfs:/tmp/mrjob']
)
Related
I am running Sqoop 1.4.7 on AWS EMR 5.21.1 and am trying to import data from a database. I have successfully been able to do this manually where I create an EMR instance with Sqoop installed via the EMR Console.
Here are the preliminary steps that I performed in order to run sqoop on EMR
Download the JDBC Driver
Move the JDBC driver to the /usr/lib/sqoop/lib directory
I was able to successfully run a sqoop import when I was sshd into an EMR cluster with these commands:
wget -O mssql-jdbc.jar https://repo1.maven.org/maven2/com/microsoft/sqlserver/mssql-jdbc/8.4.0.jre8/mssql-jdbc-8.4.0.jre8.jar
sudo mv mssql-jdbc.jar /usr/lib/sqoop/lib/
When I try to run these commands from an EMR bootstrap script however I get the error:
usr/lib/sqoop/lib/ No such file or directory
After doing some investigation I realized this is because "Bootstrap actions execute before core services, such as Hadoop or Spark, are installed", as found here
So the /usr/lib/sqoop/lib directory doesnt exist when I run my bootstrap steps.
Here are some solutions which work but they feel like work-arounds
Create the /usr/lib/sqoop/lib directory in my bootstrap script and then place the jar in it
Add the jar to this directory as an EMR step. (Turns out this this is the correct approach, look at below accepted answer)
What is the correct way of installing this JDBC driver on EMR?
The 2nd option is the correct way to do it. The documentation explains running bash scripts as an EMR step.
You can also use the jar command-runner.jar and the arguments to be
bash -c "wget -O mssql-jdbc.jar https://repo1.maven.org/maven2/com/microsoft/sqlserver/mssql-jdbc/8.4.0.jre8/mssql-jdbc-8.4.0.jre8.jar;sudo mv mssql-jdbc.jar /usr/lib/sqoop/lib/"
I have created a basic EMR cluster in AWS, and I'm trying to use the Jupyter Notebooks provided through the AWS Console. Launching the notebooks seems to work fine, and I'm also able to run basic python code in notebooks started with the pyspark kernel. Two variables are set up in the notebook: spark is a SparkSession instance, and sc is a SparkContext instance. Displaying sc yields <SparkContext master=yarn appName=livy-session-0> (the output can of course vary slightly depending on the session).
The problem arises once I perform operations that actually hit the spark machinery. For example:
sc.parallelize(list(range(10))).map(lambda x: x**2).collect()
I am no spark expert, but I believe this code should distribute the integers from 0 to 9 across the cluster, square them, and return the results in a list. Instead, I get a lengthy stack trace, mostly from the JVM, but also some python components. I believe the central part of the stack trace is the following:
py4j.protocol.Py4JJavaError: An error occurred while calling z:org.apache.spark.api.python.PythonRDD.collectAndServe.
: org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in stage 4.0 failed 4 times, most recent failure: Lost task 0.3 in stage 4.0 (TID 116, ip-XXXXXXXXXXXXX.eu-west-1.compute.internal, executor 17): java.lang.RuntimeException: Failed to run command: /usr/bin/virtualenv -p python3 --system-site-packages virtualenv_application_1586243436143_0002_0
The full stack trace is here.
A bit of digging in the AWS portal led me to log output from the nodes. stdout from one of the nodes includes the following:
The path python3 (from --python=python3) does not exist
I tried running the /usr/bin/virtualenv command on the master node manually (after logging in through), and that worked fine, but the error is of course still present after I did that.
While this error occurs most of the time, I was able to get this working in one session, where I could run several operations against the spark cluster as I was expecting.
Technical information on the cluster setup:
emr-6.0.0
Applications installed are "Ganglia 3.7.2, Spark 2.4.4, Zeppelin 0.9.0, Livy 0.6.0, JupyterHub 1.0.0, Hive 3.1.2". Hadoop is also included.
3 nodes (one of them as master), all r5a.2xlarge.
Any ideas what I'm doing wrong? Note that I am completely new to EMR and Spark.
Edit: Added the stdout log and information about running the virtualenv command manually on the master node through ssh.
I have switched to using emr-5.29.0, which seems to resolve the problem. Perhaps this is an issue with emr-6.0.0? In any case, I have a functional workaround.
The issue for me was that the virtualenv was being made on the executors with a python path that didn't exist. Pointing the executors to the right one did the job for me:
"spark.pyspark.python": "/usr/bin/python3.7"
Here is how I reconfiged the spark app at the beginning of the notebook:
{"conf":{"spark.pyspark.python": "/usr/bin/python3.7",
"spark.pyspark.virtualenv.enabled": "true",
"spark.pyspark.virtualenv.type": "native",
"spark.pyspark.virtualenv.bin.path":"/usr/bin/virtualenv"}
}
I am playing with apache-spark on aws emr, and trying to use this to set the cluster to use python3,
I use the command as the last command in a bootstrap script
sudo sed -i -e '$a\export PYSPARK_PYTHON=/usr/bin/python3' /etc/spark/conf/spark-env.sh
When I use it the cluster crashes during the bootstrap with the following error.
sed: can't read /etc/spark/conf/spark-env.sh: No such file or
directory
How should I set it to use python3 properly?
This is not a duplicate of, My issue is that the cluster is not finding the spark-env.sh file while bootstrapping, while the other question addresses the issue of the system not finding python3
In the end I did not use that script, but Used the EMR configuration file that is available on the creation stage, It gave me the proper configurations via spark_submit (in the aws gui) If you need it to be available for pyspark scripts in a more programatic way, you can use os.environ to set the pyspark python version in the python script
I have a cluster up and running. I am trying to add a step to run my code. The code itself works fine on a single instance. Only thing is I can't get it to work off S3.
aws emr add-steps --cluster-id j-XXXXX --steps Type=spark,Name=SomeSparkApp,Args=[--deploy-mode,cluster,--executor-memory,0.5g,s3://<mybucketname>/mypythonfile.py]
This is exactly what examples show I should do. What am I doing wrong?
Error I get:
Exception in thread "main" java.lang.IllegalArgumentException: Unknown/unsupported param List(--executor-memory, 0.5g, --executor-cores, 2, --primary-py-file, s3://<mybucketname>/mypythonfile.py, --class, org.apache.spark.deploy.PythonRunner)
Usage: org.apache.spark.deploy.yarn.Client [options]
Options:
--jar JAR_PATH Path to your application's JAR file (required in yarn-cluster
mode)
.
.
.
at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:121)
at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)
Command exiting with ret '1'
When I specify as this instead:
aws emr add-steps --cluster-id j-XXXXX --steps Type=spark,Name= SomeSparkApp,Args=[--executor-memory,0.5g,s3://<mybucketname>/mypythonfile.py]
I get this error instead:
Error: Only local python files are supported: Parsed arguments:
master yarn-client
deployMode client
executorMemory 0.5g
executorCores 2
EDIT: IT gets further along when I manually create the python file after SSH'ing into the cluster, and specifying as follows:
aws emr add-steps --cluster-id 'j-XXXXX' --steps Type=spark,Name= SomeSparkApp,Args=[--executor-memory,1g,/home/hadoop/mypythonfile.py]
But, not doing the job.
Any help appreciated. This is really frustrating as a well documented method on AWS's own blog here https://blogs.aws.amazon.com/bigdata/post/Tx578UTQUV7LRP/Submitting-User-Applications-with-spark-submit does not work.
I will ask, just in case, you used your correct buckets and cluster ID-s?
But anyways, I had similar problems, like I could not use --deploy-mode,cluster when reading from S3.
When I used --deploy-mode,client,--master,local[4] in the arguments, then I think it worked. But I think I still needed something different, can't remember exactly, but I resorted to a solution like this:
Firstly, I use a bootstrap action where a shell script runs the command:
aws s3 cp s3://<mybucket>/wordcount.py wordcount.py
and then I add a step to the cluster creation through the SDK in my Go application, but I can recollect this info and give you the CLI command like this:
aws emr add-steps --cluster-id j-XXXXX --steps Type=CUSTOM_JAR,Name="Spark Program",Jar="command-runner.jar",ActionOnFailure=CONTINUE,Args=["spark-submit",--master,local[4],/home/hadoop/wordcount.py,s3://<mybucket>/<inputfile.txt> s3://<mybucket>/<outputFolder>/]
I searched for days and finally discovered this thread which states
PySpark currently only supports local
files. This does not mean it only runs in local mode, however; you can
still run PySpark on any cluster manager (though only in client mode). All
this means is that your python files must be on your local file system.
Until this is supported, the straightforward workaround then is to just
copy the files to your local machine.
I created an EMR 4.0 instance in AWS with all available applications, including Spark. I did it manually, through AWS Console. I started the cluster and SSHed to the master node when it was up. There I ran pyspark. I am getting the following error when pyspark tries to create SparkContext:
2015-09-03 19:36:04,195 ERROR Thread-3 spark.SparkContext
(Logging.scala:logError(96)) - -ec2-user, access=WRITE,
inode="/user":hdfs:hadoop:drwxr-xr-x at
org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.checkFsPermission(FSPermissionChecker.java:271)
I haven't added any custom applications, nor bootstrapping and expected everything to work without errors. Not sure what's going on. Any suggestions will be greatly appreciated.
Login as the user "hadoop" (http://docs.aws.amazon.com/ElasticMapReduce/latest/ManagementGuide/emr-connect-master-node-ssh.html). It has all the proper environment and related settings for working as expected. The error you are receiving is due to logging in as "ec2-user".
I've been working with Spark on EMR this week, and found a few weird things relating to user permissions and relative paths.
It seems that running Spark from a directory which you don't 'own', as a user, is problematic. In some situations Spark (or some of the underlying Java pieces) want to create files or folders, and they think that pwd - the current directory - is the best place to do that.
Try going to the home directory
cd ~
then running pyspark.