I am currently working on a Jupyter (Lab) and PySpark 2.1.1.
I want to change spark.yarn.queue and master from a notebook. Because of the kernel spark and sc are available when I open a notebook.
Following this question, I tried
spark.conf.set("spark.yarn.queue", "my_queue")
But according to spark.sparkContext.getConf() the above line has no affect.
spark.conf.setMaster("yarn-cluster")
is not working, because there is no such a method for spark.conf.
Question: How can I change the configuration (queue and master) from a Jupyter notebook?
(Or should I set any environment variables?)
You can try to initialize spark beforehand, not in the notebook. Run this on your terminal:
export PYSPARK_DRIVER_PYTHON=jupyter
export PYSPARK_DRIVER_PYTHON_OPTS='notebook'
pyspark --master <your master> --conf <your configuration> <or any other option that pyspark supports>.
My source
Related
I'm using Pyspark from a Jupyter notebook and attempting to write a large parquet dataset to S3.
I get a 'no space left on device' error. I searched around and learned that it's because /tmp is filling up.
I want to now edit spark.local.dir to point to a directory that has space.
How can I set this parameter?
Most solutions I found suggested setting it when using spark-submit. However, I am not using spark-submit and just running it as a script from Jupyter.
Edit: I'm using Sparkmagic to work with an EMR backend.I think spark.local.dir needs to be set in the config JSON, but I am not sure how to specify it there.
I tried adding it in session_configs but it didn't work.
The answer depends on where your SparkContext comes from.
If you are starting Jupyter with pyspark:
PYSPARK_DRIVER_PYTHON='jupyter'\
PYSPARK_DRIVER_PYTHON_OPTS="notebook" \
PYSPARK_PYTHON="python" \
pyspark
then your SparkContext is already initialized when you receive your Python kernel in Jupyter. You should therefore pass a parameter to pyspark (at the end of the command above): --conf spark.local.dir=...
If you are constructing a SparkContext in Python
If you have code in your notebook like:
import pyspark
sc = pyspark.SparkContext()
then you can configure the Spark context before creating it:
import pyspark
conf = pyspark.SparkConf()
conf.set('spark.local.dir', '...')
sc = pyspark.SparkContext(conf=conf)
Configuring Spark from the command line:
It's also possible to configure Spark by editing a configuration file in bash. The file you want to edit is ${SPARK_HOME}/conf/spark-defaults.conf. You can append to it as follows (creating it if it doesn't exist):
echo 'spark.local.dir /foo/bar' >> ${SPARK_HOME}/conf/spark-defaults.conf
Now I have succeeded in running Pyspark in Jupyter in local mode by the second method as mentioned in this blog.
Here is the code:
import findspark
findspark.init()
from pyspark import SparkContext
sc = SparkContext("local", "First App")
I want to run it interactively in YARN-client mode,how can I do it?
Let's go futher,how to run in different modes,e.g.standalone mode and YARN-cluster mode.
Accrding to the Docs :
Master URLs accepts yarn parameter based on the HADOOP_CONF_DIR or YARN_CONF_DIR variable
So I can simply use:
sc = SparkContext("yarn-client", "First App")
I am trying to submit a spark application to a standalone spark(2.1.1) cluster 3 VM running in GCP from my Python 3 notebook(running in local laptop) but for some reason spark session is throwing error "StandaloneAppClient$ClientEndpoint: Failed to connect to master sparkmaster:7077".
Environment Details: IPython and Spark Master are running in one GCP VM called "sparkmaster". 3 additional GCP VMs are running Spark workers and Cassandra Clusters. I connect from my local laptop(MBP) using Chrome to GCP VM IPython notebook in "sparkmaster"
Please note that terminal works:
bin/spark-submit --packages org.apache.spark:spark-sql-kafka-0-10_2.11:2.1.1 --master spark://sparkmaster:7077 ex.py 1000
Running it from Python Notebook:
import os
os.environ["PYSPARK_SUBMIT_ARGS"] = '--packages org.apache.spark:spark-sql-kafka-0-10_2.11:2.1.1 pyspark-shell'
from pyspark.sql import SparkSession
from pyspark.sql.functions import *
spark=SparkSession.builder.master("spark://sparkmaster:7077").appName('somatic').getOrCreate() #This step works if make .master('local')
df = spark \
.readStream \
.format("kafka") \
.option("kafka.bootstrap.servers", "kafka1:9092,kafka2:9092,kafka3:9092") \
.option("subscribe", "gene") \
.load()
so far I have tried these:
I have tried to change spark master node spark-defaults.conf and spark-env.sh to add SPARK_MASTER_IP.
Tried to find the STANDALONE_SPARK_MASTER_HOST=hostname -f setting so that I can remove "-f". For some reason my spark master ui shows FQDN:7077 not hostname:7077
passed FQDN as param to .master() and os.environ["PYSPARK_SUBMIT_ARGS"]
Please let me know if you need more details.
After doing some more research I was able to resolve the conflict. It was due to a simple environment variable called SPARK_HOME. In my case it was pointing to Conda's /bin(pyspark was running from this location) whereas my spark setup was present in a diff. path. The simple fix was to add
export SPARK_HOME="/home/<<your location path>>/spark/" to .bashrc file( I want this to be attached to my profile not to the spark session)
How I have done it:
Step 1: ssh to master node in my case it was same as ipython kernel/server VM in GCP
Step 2:
cd ~
sudo nano .bashrc
scroll down to the last line and paste the below line
export SPARK_HOME="/home/your/path/to/spark-2.1.1-bin-hadoop2.7/"
ctrlX and Y and enter to save the changes
Note: I have also added few more details to the environment section for clarity.
since updating to Spark 2.3.0, tests which are run in my CI (Semaphore) fail due to a allegedly invalid spark url when creating the (local) spark context:
18/03/07 03:07:11 ERROR SparkContext: Error initializing SparkContext.
org.apache.spark.SparkException: Invalid Spark URL: spark://HeartbeatReceiver#LXC_trusty_1802-d57a40eb:44610
at org.apache.spark.rpc.RpcEndpointAddress$.apply(RpcEndpointAddress.scala:66)
at org.apache.spark.rpc.netty.NettyRpcEnv.asyncSetupEndpointRefByURI(NettyRpcEnv.scala:134)
at org.apache.spark.rpc.RpcEnv.setupEndpointRefByURI(RpcEnv.scala:101)
at org.apache.spark.rpc.RpcEnv.setupEndpointRef(RpcEnv.scala:109)
at org.apache.spark.util.RpcUtils$.makeDriverRef(RpcUtils.scala:32)
at org.apache.spark.executor.Executor.<init>(Executor.scala:155)
at org.apache.spark.scheduler.local.LocalEndpoint.<init>(LocalSchedulerBackend.scala:59)
at org.apache.spark.scheduler.local.LocalSchedulerBackend.start(LocalSchedulerBackend.scala:126)
at org.apache.spark.scheduler.TaskSchedulerImpl.start(TaskSchedulerImpl.scala:164)
at org.apache.spark.SparkContext.<init>(SparkContext.scala:500)
at org.apache.spark.SparkContext$.getOrCreate(SparkContext.scala:2486)
at org.apache.spark.sql.SparkSession$Builder$$anonfun$7.apply(SparkSession.scala:930)
at org.apache.spark.sql.SparkSession$Builder$$anonfun$7.apply(SparkSession.scala:921)
at scala.Option.getOrElse(Option.scala:121)
at org.apache.spark.sql.SparkSession$Builder.getOrCreate(SparkSession.scala:921)
The spark session is created as following:
val sparkSession: SparkSession = SparkSession
.builder
.appName(s"LocalTestSparkSession")
.config("spark.broadcast.compress", "false")
.config("spark.shuffle.compress", "false")
.config("spark.shuffle.spill.compress", "false")
.master("local[3]")
.getOrCreate
Before updating to Spark 2.3.0, no problems were encountered in version 2.2.1 and 2.1.0. Also, running the tests locally works fine.
Change the SPARK_LOCAL_HOSTNAME to localhost and try.
export SPARK_LOCAL_HOSTNAME=localhost
This has been resolved by setting sparkSession config "spark.driver.host" to the IP address.
It seems that this change is required from 2.3 onwards.
If you don't want to change the environment variable, you can change the code to add the config in the SparkSession builder (like Hanisha said above).
In PySpark:
spark = SparkSession.builder.config("spark.driver.host", "localhost").getOrCreate()
As mentioned in above answers, You need to change SPARK_LOCAL_HOSTNAME to localhost. In windows, you have to use SET command, SET SPARK_LOCAL_HOSTNAME=localhost
but this SET command is temporary. you may have to run it again and again in every new terminal. but instead, you can use SETX command, which is permanent.
SETX SPARK_LOCAL_HOSTNAME localhost
You can type above command in any place. just open a command prompt and run above command. Notice that unlike SET command, SETX command do not allow equation mark. you need to separate environment variable and the value by a Space.
if Success, you will see a message like "SUCCESS: Specified value was saved"
you can also verify that your variable is successfully added by just typing SET in a different command prompt. (or type SET s , which gives variables, starting with the letter 'S'). you can see that SPARK_LOCAL_HOSTNAME=localhost in results, which will not happen if you use SET command instead of SETX
Change your hostname to have NO underscore.
spark://HeartbeatReceiver#LXC_trusty_1802-d57a40eb:44610 to spark://HeartbeatReceiver#LXCtrusty1802d57a40eb:44610
Ubuntu AS root
#hostnamectl status
#hostnamectl --static set-hostname LXCtrusty1802d57a40eb
#nano /etc/hosts
127.0.0.1 LXCtrusty1802d57a40eb
#reboot
Try to run Spark locally, with as many worker threads as logical cores on your machine :
.master("local[*]")
I would like to complement #Prakash Annadurai answer by saying:
If you want to make the variable settlement last after exiting the terminal, add it to your shell profile (e.g. ~/.bash_profile) with the same command:
export SPARK_LOCAL_HOSTNAME=localhost
For anyone working in Jupyter Notebook. Adding %env SPARK_LOCAL_HOSTNAME=localhost to the very beginning of the cell solved it for me. Like so:
%env SPARK_LOCAL_HOSTNAME=localhost
import findspark
findspark.init()
from pyspark import SparkConf, SparkContext
conf = SparkConf().setMaster("local").setAppName("Test")
sc = SparkContext(conf = conf)
Setting .config("spark.driver.host", "localhost") fixed the issue for me.
SparkSession spark = SparkSession
.builder()
.config("spark.master", "local")
.config("fs.s3a.aws.credentials.provider", "org.apache.hadoop.fs.s3a.TemporaryAWSCredentialsProvider")
.config("spark.hadoop.fs.s3a.buffer.dir", "/tmp")
.config("spark.driver.memory", "2048m")
.config("spark.executor.memory", "2048m")
.config("spark.driver.bindAddress", "127.0.0.1")
.config("spark.driver.host", "localhost")
.getOrCreate();
I have a machine with Hadoop and Spark installed. Below is my current environment.
python3.6
spark1.5.2
Hadoop 2.7.1.2.3.6.0-3796
I was trying to connect jupyter notebook to connect to spark by building up ipython kernel.
2 new files written.
/root/.ipython/profile_pyspark/ipython_notebook_config.py
/root/.ipython/profile_pyspark/startup/00-pyspark-setup.py
/root/anaconda3/share/jupyter/kernels/pyspark/kernel.json
kernel.json
{
"display_name": "PySpark (Spark 2.0.0)",
"language": "python",
"argv": [
"/root/anaconda3/bin/python3",
"-m",
"ipykernel",
"--profile=pyspark"
],
"env": {
"CAPTURE_STANDARD_OUT": "true",
"CAPTURE_STANDARD_ERR": "true",
"SEND_EMPTY_OUTPUT": "false",
"PYSPARK_PYTHON" : "/root/anaconda3/bin/python3",
"SPARK_HOME": "/usr/hdp/current/spark-client/"
}
}
00-pyspark-setup.py
import os
import sys
os.environ["PYSPARK_PYTHON"] = "/root/anaconda3/bin/python"
os.environ["SPARK_HOME"] = "/usr/hdp/current/spark-client"
os.environ["PYLIB"] = os.environ["SPARK_HOME"] + "/python/lib"
spark_home = os.environ.get('SPARK_HOME', None)
sys.path.insert(0, os.environ["PYLIB"] +"/py4j-0.8.2.1-src.zip")
sys.path.insert(0, os.environ["PYLIB"] +"/pyspark.zip")
exec(open(os.path.join(spark_home, 'python/pyspark/shell.py')).read())
ipython_notebook_config.py
c = get_config()
c.NotebookApp.port = 80
Then, when i run the following
jupyter notebook --profile=pyspark
The notebook is running well. Then, i change kernel to 'PySpark (Spark 2.0.0)', and it suppose to use the 'sc' spark context . However, when i type 'sc', it does not shows anything.
So, since sc cannot be initiliazed, if i want to run the following, it failed!
nums = sc.parallelize(xrange(1000000))
Can anybody help me how to configure jupyter notebook to talk to Spark?
Just FYI, python 3.6 isnt supported until version spark 2.1.1. See JIRA https://issues.apache.org/jira/browse/SPARK-19019
There is a number of issues with your question...
1) On top of the answer by Punskr above - Spark 1.5 only works with Python 2; Python 3 support was introduced in Spark 2.0.
2) Even if you switch to Python 2 or upgrade Spark, you will still need to import the relevant modules of Pyspark and initialize the sc variable manually in the notebook
3) You also seem to use an old version of Jupyter, since the profiles functionality is not available in Jupyter >= 4.
To initialise sc "automatically" in Jupyter >=4, see my answer here.
You can make a few environment changes to have pyspark default ipython or a jupyter notebook.
Put the following in your ~/.bashrc
export PYSPARK_PYTHON=python3 ## for python3
export PYSPARK_DRIVER_PYTHON=ipython
export PYSPARK_DRIVER_PYTHON_OPTS="notebook --no-browser --port=7000"
See: pyspark on GitHub
Next, run source ~/.bashrc
Then, when you launch pyspark (or with YARN) it will open up a server for you to connect to.
On a local terminal that has ssh capabilities, run
ssh -N -f -L localhost:8000:localhost:7000 <username>#<host>
If you're on Windows, I recommend MobaXterm or Cygwin.
Open up a web browser, and enter the address localhost:8000 to tunnel into your notebook with Spark
Now some precautions, I've never tried this with Python 3 so this may or may not work for you. Regardless, you should really be using Python 2 on Spark 1.5. My company uses Spark 1.5 as well, as NO ONE uses Python 3 because of it.
Update:
Per #desertnaut's comments, setting
export PYSPARK_DRIVER_PYTHON=ipython
may cause issues if the user ever needs to use spark-submit. A work around, if you want to have both notebooks and spark-submit available is to create two new environment variables. Here is an example of what you may create
export PYSPARK_PYTHON=python3 ## for python3
export ipyspark='PYSPARK_DRIVER_PYTHON=ipython pyspark'
export pynb='PYSPARK_DRIVER_PYTHON_OPTS="notebook --no-browser --port=7000"'
where ipyspark and pynb are new commands on a bash terminal.