I use this command to summit spark application to yarn cluster
export YARN_CONF_DIR=conf
bin/spark-submit --class "Mining"
--master yarn-cluster
--executor-memory 512m ./target/scala-2.10/mining-assembly-0.1.jar
In Web UI, it stuck on UNDEFINED
In console, it stuck to
<code>14/11/12 16:37:55 INFO yarn.Client: Application report from ASM:
application identifier: application_1415704754709_0017
appId: 17
clientToAMToken: null
appDiagnostics:
appMasterHost: example.com
appQueue: default
appMasterRpcPort: 0
appStartTime: 1415784586000
yarnAppState: RUNNING
distributedFinalState: UNDEFINED
appTrackingUrl: http://example.com:8088/proxy/application_1415704754709_0017/
appUser: rain
</code>
Update:
Dive into Logs for container in Web UI http://example.com:8042/node/containerlogs/container_1415704754709_0017_01_000001/rain/stderr/?start=0, I found this
14/11/12 02:11:47 WARN YarnClusterScheduler: Initial job has not accepted
any resources; check your cluster UI to ensure that workers are registered
and have sufficient memory
14/11/12 02:11:47 DEBUG Client: IPC Client (1211012646) connection to
spark.mvs.vn/192.168.64.142:8030 from rain sending #24418
14/11/12 02:11:47 DEBUG Client: IPC Client (1211012646) connection to
spark.mvs.vn/192.168.64.142:8030 from rain got value #24418
I found this problem have had solution here http://hortonworks.com/hadoop-tutorial/using-apache-spark-hdp/
The Hadoop cluster must have sufficient memory for the request.
For example, submitting the following job with 1GB memory allocated for
executor and Spark driver fails with the above error in the HDP 2.1 Sandbox.
Reduce the memory asked for the executor and the Spark driver to 512m and
re-start the cluster.
I'm trying this solution and hopefully it will work.
Solutions
Finally I found that it caused by memory problem
It worked when I change yarn.nodemanager.resource.memory-mb to 3072 (its value was 2048) in Web UI of interface and restarted cluster.
I'm very happy to see this
With 3GB in yarn nodemanager, my summit is
bin/spark-submit
--class "Mining"
--master yarn-cluster
--executor-memory 512m
--driver-memory 512m
--num-executors 2
--executor-cores 1
./target/scala-2.10/mining-assembly-0.1.jar`
Related
When I submit this command, my job failed with error "Container is running beyond physical memory limits".
spark-submit --master yarn --deploy-mode cluster --executor-memory 5G --total-executor-cores 30 --num-executors 15 --conf spark.yarn.executor.memoryOverhead=1000
But adding the parameter: --driver-memory to 5GB (or upper), the job ends without error.
spark-submit --master yarn --deploy-mode cluster --executor-memory 5G --total executor-cores 30 --num-executors 15 --driver-memory 5G --conf spark.yarn.executor.memoryOverhead=1000
Cluster info: 6 nodes with 120GB of Memory. YARN Container Memory Minimum: 1GB
The question is: what is the difference in using or not this parameter?
If increasing the driver memory is helping you to successfully complete the job then it means that driver is having lots of data coming into it from executors. Typically, the driver program is responsible for collecting results back from each executor after the tasks are executed. So, in your case it seems that increasing the driver memory helped to store more results back into the driver memory.
If you read the some points on executor memory, driver memory and the way Driver interacts with executors then you will get better clarity on the situation you are in.
Hope it helps to some extent.
I am running Spark on YARN in client mode, so I expect that YARN will allocate containers only for the executors. Yet, from what I am seeing, it seems like a container is also allocated for the driver, and I don't get as many executors as I was expecting.
I am running spark submit on the master node. Parameters are as follows:
sudo spark-submit --class ... \
--conf spark.master=yarn \
--conf spark.submit.deployMode=client \
--conf spark.yarn.am.cores=2 \
--conf spark.yarn.am.memory=8G \
--conf spark.executor.instances=5 \
--conf spark.executor.cores=3 \
--conf spark.executor.memory=10G \
--conf spark.dynamicAllocation.enabled=false \
While running this application, Spark UI's Executors page shows 1 driver and 4 executors (5 entries in total). I would expect 5, not 4 executors.
At the same time, YARN UI's Nodes tab shows that on the node that isn't actually used (at least according to Spark UI's Executors page...) there's a container allocated, using 9GB of memory. The rest of the nodes have containers running on them, 11GB of memory each.
Because in my Spark Submit the driver has 2GB less memory than executors, I think that the 9GB container allocated by YARN is for the driver.
Why is this extra container allocated? How can i prevent this?
Spark UI:
YARN UI:
Update after answer by Igor Dvorzhak
I was falsely assuming that the AM will run on the master node, and that it will contain the driver app (so setting spark.yarn.am.* settings will relate to the driver process).
So I've made the following changes:
set the spark.yarn.am.* settings to defaults (512m of memory, 1 core)
set the driver memory through spark.driver.memory to 8g
did not try to set driver cores at all, since it is only valid for cluster mode
Because AM on default settings takes up 512m + 384m of overhead, its container fits into the spare 1GB of free memory on a worker node.
Spark gets the 5 executors it requested, and the driver memory is appropriate to the 8g setting. All works as expected now.
Spark UI:
YARN UI:
Extra container is allocated for YARN application master:
In client mode, the driver runs in the client process, and the application master is only used for requesting resources from YARN.
Even though in client mode driver runs in the client process, YARN application master is still running on YARN and requires container allocation.
There are no way to prevent container allocation for YARN application master.
For reference, similar question asked time ago: Resource Allocation with Spark and Yarn.
You can specify the driver memory and number of executors in spark submit as below.
spark-submit --jars..... --master yarn --deploy-mode cluster --driver-memory 2g --driver-cores 4 --num-executors 5 --executor-memory 10G --executor-cores 3
Hope it helps you.
I have Spark installed, but when I launch, there is always only one executor allocated to the application (and that is the driver one). I’ve tried everything, but haven’t been able to find out why this is happening.
Here’s the command I used to launch, to give you an idea of all the parameters:
dcos spark run --submit-args='--class <class-name> --executor-memory 6g --total-executor-cores 32 --driver-memory 6g <jar-file-source> <application-command-line-params>
I'm trying to run a Spark application in a Mesos cluster where I have one master and one slave. The slave has 8GB RAM assigned for Mesos. The master is running the Spark Mesos Dispatcher.
I use the following command to submit a Spark application (which is a streaming application).
spark-submit --master mesos://mesos-master:7077 --class com.verifone.media.ums.scheduling.spark.SparkBootstrapper --deploy-mode cluster scheduling-spark-0.5.jar
And I see the following output which shows its successfully submitted.
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
15/09/01 12:52:38 INFO RestSubmissionClient: Submitting a request to launch an application in mesos://mesos-master:7077.
15/09/01 12:52:39 INFO RestSubmissionClient: Submission successfully created as driver-20150901072239-0002. Polling submission state...
15/09/01 12:52:39 INFO RestSubmissionClient: Submitting a request for the status of submission driver-20150901072239-0002 in mesos://mesos-master:7077.
15/09/01 12:52:39 INFO RestSubmissionClient: State of driver driver-20150901072239-0002 is now QUEUED.
15/09/01 12:52:40 INFO RestSubmissionClient: Server responded with CreateSubmissionResponse:
{
"action" : "CreateSubmissionResponse",
"serverSparkVersion" : "1.4.1",
"submissionId" : "driver-20150901072239-0002",
"success" : true
}
However, this fails in Mesos, and when I look at the Spark Cluster UI, I see the following message.
task_id { value: "driver-20150901070957-0001" } state: TASK_FAILED message: "" slave_id { value: "20150831-082639-167881920-5050-4116-S6" } timestamp: 1.441091399975446E9 source: SOURCE_SLAVE reason: REASON_MEMORY_LIMIT 11: "\305-^E\377)N\327\277\361:\351\fm\215\312"
Seems like it is related to memory, but I'm not sure whether I have to configure something here to get this working.
UPDATE
I looked at the mesos logs in the slave, and I see the following message.
E0901 07:56:26.086618 1284 fetcher.cpp:515] Failed to run mesos-fetcher: Failed to fetch all URIs for container '33183181-e91b-4012-9e21-baa37485e755' with exit status: 256
So I thought that this could be because of the Spark Executor URL, so I modified the spark-submit to be as follows and increased memory for both driver and slave, but still I see the same error.
spark-submit \
--master mesos://mesos-master:7077 \
--class com.verifone.media.ums.scheduling.spark.SparkBootstrapper \
--deploy-mode cluster \
--driver-memory 1G \
--executor-memory 4G \
--conf spark.executor.uri=http://d3kbcqa49mib13.cloudfront.net/spark-1.4.1-bin-hadoop2.6.tgz \
scheduling-spark-0.5.jar
UPDATE 2
I went past this point by following #hartem's advice (see comments). Tasks are running now, but still, actual Spark application does not run in the cluster. When I look at the logs I see the following. After the last line, seems that Spark does not proceed any further.
15/09/01 10:33:41 INFO SparkContext: Added JAR file:/tmp/mesos/slaves/20150831-082639-167881920-5050-4116-S8/frameworks/20150831-082639-167881920-5050-4116-0004/executors/driver-20150901103327-0002/runs/47339c12-fb78-43d6-bc8a-958dd94d0ccf/spark-1.4.1-bin-hadoop2.6/../scheduling-spark-0.5.jar at http://192.172.1.31:33666/jars/scheduling-spark-0.5.jar with timestamp 1441103621639
I0901 10:33:41.728466 4375 sched.cpp:157] Version: 0.23.0
I0901 10:33:41.730764 4383 sched.cpp:254] New master detected at master#192.172.1.10:7077
I0901 10:33:41.730908 4383 sched.cpp:264] No credentials provided. Attempting to register without authentication
I had similar issue problem was slave could not find the required jar for running the class file(SparkPi). So i gave the http URL of the jar it worked, it requires jar to be placed in distributed system not on local file system.
/home/centos/spark-1.6.1-bin-hadoop2.6/bin/spark-submit \
--name SparkPiTestApp \
--class org.apache.spark.examples.SparkPi \
--master mesos://xxxxxxx:7077 \
--deploy-mode cluster \
--executor-memory 5G --total-executor-cores 30 \
http://downloads.mesosphere.com.s3.amazonaws.com/assets/spark/spark-examples_2.10-1.4.0-SNAPSHOT.jar 100
Could you please do export GLOG_v=1 before launching the slave and see if there is anything interesting in the slave log? I would also look for stdout and stderr files under the slave working directory and see if they contain any clues.
I have one Spark job which runs fine locally with less data but when I schedule it on YARN to execute I keep on getting the following error and slowly all executors get removed from UI and my job fails
15/07/30 10:18:13 ERROR cluster.YarnScheduler: Lost executor 8 on myhost1.com: remote Rpc client disassociated
15/07/30 10:18:13 ERROR cluster.YarnScheduler: Lost executor 6 on myhost2.com: remote Rpc client disassociated
I use the following command to schedule Spark job in yarn-client mode
./spark-submit --class com.xyz.MySpark --conf "spark.executor.extraJavaOptions=-XX:MaxPermSize=512M" --driver-java-options -XX:MaxPermSize=512m --driver-memory 3g --master yarn-client --executor-memory 2G --executor-cores 8 --num-executors 12 /home/myuser/myspark-1.0.jar
What is the problem here? I am new to Spark.
I had a very similar problem. I had many executors being lost no matter how much memory we allocated to them.
The solution if you're using yarn was to set --conf spark.yarn.executor.memoryOverhead=600, alternatively if your cluster uses mesos you can try --conf spark.mesos.executor.memoryOverhead=600 instead.
In spark 2.3.1+ the configuration option is now --conf spark.yarn.executor.memoryOverhead=600
It seems like we were not leaving sufficient memory for YARN itself and containers were being killed because of it. After setting that we've had different out of memory errors, but not the same lost executor problem.
You can follow this AWS post to calculate memory overhead (and other spark configs to tune): best-practices-for-successfully-managing-memory-for-apache-spark-applications-on-amazon-emr
When I had the same issue, deleting logs and free up more hdfs space worked.