I have researched this for a significant amount of time and find answers that seem to be for a slightly different question than mine.
UPDATE: Spark docs say the Driver runs on a cluster Worker in deployMode: cluster. This does not seem to be true when you don't use spark-submit
My Spark 2.3.3 cluster is running fine. I see the GUI on “http://master-address:8080", there are 2 idle workers, as configured.
I have a Scala application that creates a context and starts a Job. I do not use spark-submit, I start the Job programmatically and this is where many answers diverge from my question.
In "my-app" I create a new SparkConf, with the following code (slightly abbreviated):
conf.setAppName(“my-job")
conf.setMaster(“spark://master-address:7077”)
conf.set(“deployMode”, “cluster”)
// other settings like driver and executor memory requests
// the driver and executor memory requests are for all mem on the slaves, more than
// mem available on the launching machine with “my-app"
val jars = listJars(“/path/to/lib")
conf.setJars(jars)
…
When I launch the job I see 2 executors running on the 2 nodes/workers/slaves. The logs show their IP address and calls them executor 0 and 1.
With a Yarn cluster I would expect the “Driver" to run on/in the Yarn Master but I am using the Spark Standalone Master, where is the Driver part of the Job running? If it runs on a random worker or elsewhere, is there a way to find it from logs
Where is my Spark Driver executing? Does deployMode = cluster work when not using spark-submit? Evidence shows a cluster with one master (on the same machine as executor 0) and 2 Workers. It also show identical memory usage on both Workers during the job. From logs I know both Workers are running Executors. Where is the Driver?
The “Driver” creates and broadcasts some large data structures so the need for an answer is more critical than with more typical tiny Drivers.
Where is the driver running? How do I find it given logs and monitoring? I can't reconcile what I see with the docs, they contradict each other.
This is answered by the official documentation:
In cluster mode, however, the driver is launched from one of the Worker processes inside the cluster, and the client process exits as soon as it fulfills its responsibility of submitting the application without waiting for the application to finish.
In other words driver uses arbitrary worker node, hence it it is likely to co-locate with one on the executors, on such small cluster. And to anticipate the follow-up question - this behavior is not configurable - you just have to make sure that the cluster has capacity to start both required executors, and the driver with it's requested memory and cores.
Related
I have three Ubuntu VMs (clones) in my local machine which i wanted to use to make a simple cluster. One VM to be used as a master and the other two as slaves. I can ssh every VM from every other one succesfully and i have the ip's of the two slaves in the conf/slaves file of the master and the master's ip in the spark-env.sh of every VM.When I run
start-slave.sh spark://master-ip:7077
from the slaves,they appear in the spark UI. But when i try to run things in parallel i always get the message about the resources. For testing code i use the scala shell
spark-shell --master://master-ip:7077 and sc.parallelize(1 until 10000).count.
Do You mean that warn: WARN TaskSchedulerImpl: Initial job has not accepted any resources; check your cluster ui to ensure that workers are registered and have sufficient memory
This message will pop up any time an application is requesting more resources from the cluster than the cluster can currently provide.
Spark is only looking for two things: Cores and Ram. Cores represents the number of open executor slots that your cluster provides for execution. Ram refers to the amount of free Ram required on any worker running your application.
Note for both of these resources the maximum value is not your System's max, it is the max as set by the your Spark configuration.
If you need to run multiple Spark apps simultaneously then you’ll need to adjust the amount of cores being used by each app.
If you are working with applications on the same node you need to assign cores to each application to make them work in parallel: ResourceScheduling
If you use VMs (as in your situation): assign only one core to each VM
when you first create it or whatever relevant to your system
resource capacity as by now spark request 4 cores for each * 2 VMs = 8 core which you don't have.
This is a tutorial i find that could help you: Install Spark on Ubuntu: Standalone Cluster Mode
Further Reading: common-spark-troubleshooting
I found some code to start spark locally with:
val conf = new SparkConf().setAppName("test").setMaster("local[*]")
val ctx = new SparkContext(conf)
What does the [*] mean?
From the doc:
./bin/spark-shell --master local[2]
The --master option specifies the master URL for a distributed
cluster, or local to run locally with one thread, or local[N] to run
locally with N threads. You should start by using local for testing.
And from here:
local[*] Run Spark locally with as many worker threads as logical
cores on your machine.
Master URL Meaning
local : Run Spark locally with one worker thread (i.e. no parallelism at all).
local[K] : Run Spark locally with K worker threads (ideally, set this to the number of cores on your machine).
local[K,F] : Run Spark locally with K worker threads and F maxFailures (see spark.task.maxFailures for an explanation of this variable)
local[*] : Run Spark locally with as many worker threads as logical cores on your machine.
local[*,F] : Run Spark locally with as many worker threads as logical cores on your machine and F maxFailures.
spark://HOST:PORT : Connect to the given Spark standalone cluster master. The port must be whichever one your master is configured to use, which is 7077 by default.
spark://HOST1:PORT1,HOST2:PORT2 : Connect to the given Spark standalone cluster with standby masters with Zookeeper. The list must have all the master hosts in the high availability cluster set up with Zookeeper. The port must be whichever each master is configured to use, which is 7077 by default.
mesos://HOST:PORT : Connect to the given Mesos cluster. The port must be whichever you have configured to use, which is 5050 by default. Or, for a Mesos cluster using ZooKeeper, use mesos://zk://.... To submit with --deploy-mode cluster, the HOST:PORT should be configured to connect to the MesosClusterDispatcher.
yarn : Connect to a YARN cluster in client or cluster mode depending on the value of --deploy-mode. The cluster location will be found based on the HADOOP_CONF_DIR or YARN_CONF_DIR variable.
https://spark.apache.org/docs/latest/submitting-applications.html
Some additional Info
Do not run Spark Streaming programs locally with master configured as "local" or "local[ 1]". This allocates only one CPU for tasks and if a receiver is running on it, there is no resource left to process the received data. Use at least "local[ 2]" to have more cores.
From -Learning Spark: Lightning-Fast Big Data Analysis
Master URL
You can run Spark in local mode using local, local[n] or the most general local[*] for the master URL.
The URL says how many threads can be used in total:
local uses 1 thread only.
local[n] uses n threads.
local[*] uses as many threads as the number of processors available to the Java virtual machine (it uses Runtime.getRuntime.availableProcessors() to know the number).
local[N, maxFailures] (called local-with-retries) with N being * or the number of threads to use (as explained above) and maxFailures being the value of spark.task.maxFailures.
You can run Spark in local mode using local, local[n] or the most general local[*] for the master URL.
The URL says how many threads can be used in total:-
local uses 1 thread only.
local[n] uses n threads.
local[*] uses as many threads as your spark local machine have, where you are running your application.
you can check by lscpu in your Linux machine
[ie#mapr2 ~]$ lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 56
On-line CPU(s) list: 0-55
Thread(s) per core: 2
if your machine has 56 cores means CPU then your spark jobs will be partitioned in 56 part.
NOTE:- there may be the case that in your spark cluster the spark-defaults.conf file has limited the partition value with the default value (like 10 or else) then your partitioned will be the same as default value has been set in config.
local[N, maxFailures] (called local-with-retries) with N being * or the number of threads to use (as explained above) and maxFailures being the value of spark.task.maxFailures.
without * spark will use single thread.
With * spark will use all the available threads the run this program
Installed spark cluster on standalone mode with 2 nodes on first node there is spark master running and on another node spark worker. When i try to run spark shell on worker node with word count code it runs fine but when i try to run spark shell on the master node it gives following output :
WARN scheduler.TaskSchedulerImpl: Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources
Executor is not triggered to run the job. Even though there is worker available to spark master its giving such a problem . Any help is appriciated , thanks
You use client deploy mode so the best bet is that executor nodes cannot connect to the driver port on the local machine. It could be firewall issue or problem with advertised IP / hostname. Please make sure that:
spark.driver.bindAddress
spark.driver.host
spark.driver.port
use expected values. Please refer to the networking section of Spark documentation.
Less likely it is a lack of resources. Please check if you don't request more resources than provided by workers.
I have a virtual machine in which a spark-2.0.0-bin-hadoop2.7 in standalone mode is installed.
I ran ./sbin/start-all.sh to run the master and the slave.
When I do ./bin/spark-shell --master spark://192.168.43.27:7077 --driver-memory 600m --executor-memory 600m --executor-cores 1 in the machine itself the task's status is RUNNING and I am able to compute code in spark shell.
When I do exactly the same command from another machine in the network, the status is "RUNNING" again, but the spark-shell throws WARN TaskSchedulerImpl: Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources. I guess the problem is not directly related to resources because the same command works in the virtual machine itself, but not when it comes from other machines.
I checked most of the topics related to this error and none of them solved my problem. I even disabled firewall with sudo ufw disable just to make sure but no success (based on this link) which suggests:
Disable Firewall on the client : This was the solution that worked for me. Since I was working on a prototype in-house code, I disabled the firewall on the client node. For some reason the worker nodes, were not able to talk back to the client for me. For production purposes, you would want to open-up certain number of ports required.
There are two known reasons for this:
Your application requires more resources (cores, memory) than allocated. Increasing worker cores and memory should solve it. Most other answers focus on this.
Where less known, the firewall is blocking the communication between master and workers. This could happen especially you are using cloud service. According to Spark Security, besides the standard 8080, 8081, 7077, 4040 ports, you also need to make sure the master and worker can communicate via the SPARK_WORKER_PORT, spark.driver.port and spark.blockManager.port; the latter three are used by submitting jobs and are randomly assigned by the program (if left unconfigured). You may try to open all ports to run a quick test.
Add an example of #Fountaine007's first bullet.
I ran into the same issue and it's because the allocated vcores is less than the application's expectation.
For my specific scenario, I increased the value of yarn.nodemanager.resource.cpu-vcores under $HADOOP_HOME/etc/hadoop/yarn-site.xml.
For memory related issue, you may also need to modify yarn.nodemanager.resource.memory-mb.
I had a question regarding this image in a tutorial I was following. So based on this image in a yarn based architecture does the execution of a spark application look something like this:
First you have a driver which is running on a client node or some data node. In this driver (similar to a driver in java?) consists of your code (written in java, python, scala, etc.) that you submit to the Spark Context. Then that spark context represents the connection to HDFS and submits your request to the Resource manager in the Hadoop ecosystem. Then the resource manager communicates with the Name node to figure out which data nodes in the cluster contain the information the client node asked for. The spark context will also put a executor on the worker node that will run the tasks. Then the node manager will start the executor which will run the tasks given to it by the Spark Context and will return back the data the client asked for from the HDFS to the driver.
Is the above interpretation correct?
Also would a driver send out three executors to each data node to retrieve the data from the HDFS, since the data in HDFS is replicated 3 times on various data nodes?
Your interpretation is close to reality but it seems that you are a bit confused on some points.
Let's see if I can make this more clear to you.
Let's say that you have the word count example in Scala.
object WordCount {
def main(args: Array[String]) {
val inputFile = args(0)
val outputFile = args(1)
val conf = new SparkConf().setAppName("wordCount")
val sc = new SparkContext(conf)
val input = sc.textFile(inputFile)
val words = input.flatMap(line => line.split(" "))
val counts = words.map(word => (word, 1)).reduceByKey{case (x, y) => x + y}
counts.saveAsTextFile(outputFile)
}
}
In every spark job you have an initialisation step where you create a SparkContext object providing some configuration like the appname and the master, then you read a inputFile, you process it and you save the result of your processing on disk. All this code is running in the Driver except for the anonymous functions that make the actual processing (functions passed to .flatMap, .map and reduceByKey) and the I/O functions textFile and saveAsTextFile which are running remotely on the cluster.
Here the DRIVER is the name that is given to that part of the program running locally on the same node where you submit your code with spark-submit (in your picture is called Client Node). You can submit your code from any machine (either ClientNode, WorderNode or even MasterNode) as long as you have spark-submit and network access to your YARN cluster. For simplicity I will assume that the Client node is your laptop and the Yarn cluster is made of remote machines.
For simplicity I will leave out of this picture Zookeeper since it is used to provide High availability to HDFS and it is not involved in running a spark application. I have to mention that Yarn Resource Manager and HDFS Namenode are roles in Yarn and HDFS (actually they are processes running inside a JVM) and they could live on the same master node or on separate machines. Even Yarn Node managers and Data Nodes are only roles but they usually live on the same machine to provide data locality (processing close to where data are stored).
When you submit your application you first contact the Resource Manager that together with the NameNode try to find Worker nodes available where to run your spark tasks. In order to take advantage of the data locality principle, the Resource Manager will prefer worker nodes that stores on the same machine HDFS blocks (any of the 3 replicas for each block) for the file that you have to process. If no worker nodes with those blocks is available it will use any other worker node. In this case since data will not be available locally, HDFS blocks has to be moved over the network from any of the Data nodes to the node manager running the spark task. This process is done for each block that made your file, so some blocks could be found locally, some have to moved.
When the ResourceManager find a worker node available it will contact the NodeManager on that node and ask it to create an a Yarn Container (JVM) where to run a spark executor. In other cluster modes (Mesos or Standalone) you won't have a Yarn container but the concept of spark executor is the same. A spark executor is running as a JVM and can run multiple tasks.
The Driver running on the client node and the tasks running on spark executors keep communicating in order to run your job. If the driver is running on your laptop and your laptop crash, you will loose the connection to the tasks and your job will fail. That is why when spark is running in a Yarn cluster you can specify if you want to run your driver on your laptop "--deploy-mode=client" or on the yarn cluster as another yarn container "--deploy-mode=cluster". For more details look at spark-submit