> On 5 Oct 2015, at 16:48, Alex Rovner wrote:
>
> Hey Steve,
>
> Are you referring to the 1.5 version of the history server?
>
Yes. I should warn, however, that there's no guarantee that a history server
running the 1.4 code will handle the histories of a 1.5+
YARN 2.7.1 (running on the cluster) was built with Java 1.8, I assume.
Have you used the following command to retrieve / inspect logs ?
yarn logs -applicationId
Cheers
On Mon, Oct 5, 2015 at 8:41 AM, mvle wrote:
> Hi,
>
> I have successfully run pyspark on Spark 1.5.1 on YARN
lto:user@spark.apache.org>"
Subject: Re: Spark on YARN / aws - executor lost on node restart
Hi guys,
Digging up this question after spending some more time trying to replicate it.
It seems to be an issue with the YARN – spark integration, wondering if there
is a bug already tracking this?
I think you need to increase the memory size of executor through command
arguments "--executor-memory", or configuration "spark.executor.memory".
Also yarn.scheduler.maximum-allocation-mb in Yarn side if necessary.
Thanks
Saisai
On Mon, Sep 21, 2015 at 5:13 PM, Alexander Pivovarov
The warning your seeing in Spark is no issue. The scratch space lives
inside the heap, so it'll never result in YARN killing the container by
itself. The issue is that Spark is using some off-heap space on top of
that.
You'll need to bump the spark.yarn.executor.memoryOverhead property to give
I repartitioned input RDD from 4,800 to 24,000 partitions
After that the stage (24000 tasks) was done in 22 min on 100 boxes
Shuffle read/write: 905 GB / 710 GB
Task Metrics (Dur/GC/Read/Write)
Min: 7s/1s/38MB/30MB
Med: 22s/9s/38MB/30MB
Max:1.8min/1.6min/38MB/30MB
On Mon, Sep 21, 2015 at 5:55
I noticed that some executors have issue with scratch space.
I see the following in yarn app container stderr around the time when yarn
killed the executor because it uses too much memory.
-- App container stderr --
15/09/21 21:43:22 WARN storage.MemoryStore: Not enough space to cache
rdd_6_346
Hi Nick/Igor,
Any solution for this ?
Even I am having the same issue and copying jar to each executor is not
feasible if we use lot of jars.
Thanks,
Vipul
Hi guys,
Digging up this question after spending some more time trying to replicate it.
It seems to be an issue with the YARN – spark integration, wondering if there
is a bug already tracking this?
If I just kill the process on the machine, YARN detects the container is dead
and the spark
Task set is a set of tasks within one stage.
Executor will be killed when it is idle for a period of time (default is
60s). The problem you mentioned is bug, scheduler should not allocate tasks
on this to-be killed executors. I think it is fixed in 1.5.
Thanks
Saisai
On Thu, Sep 17, 2015 at
YARN will never kill processes for being unresponsive.
It may kill processes for occupying more memory than it allows. To get
around this, you can either bump spark.yarn.executor.memoryOverhead or turn
off the memory checks entirely with yarn.nodemanager.pmem-check-enabled.
-Sandy
On Tue, Sep
as a starting point, attach your stacktrace...
ps: look for duplicates in your classpath, maybe you include another jar
with same class
On 8 September 2015 at 06:38, Nicholas R. Peterson
wrote:
> I'm trying to run a Spark 1.4.1 job on my CDH5.4 cluster, through Yarn.
>
Thans, Igor; I've got it running again right now, and can attach the stack
trace when it finishes.
In the mean time, I've noticed something interesting: in the Spark UI, the
application jar that I submit is not being included on the classpath. It
has been successfully uploaded to the nodes -- in
Those settings seem reasonable to me.
Are you observing performance that's worse than you would expect?
-Sandy
On Mon, Sep 7, 2015 at 11:22 AM, Alexander Pivovarov
wrote:
> Hi Sandy
>
> Thank you for your reply
> Currently we use r3.2xlarge boxes (vCPU: 8, Mem: 61 GiB)
>
Yes, the jar contains the class:
$ jar -tf lumiata-evaluation-assembly-1.0.jar | grep 2028/Document/Document
com/i2028/Document/Document$1.class
com/i2028/Document/Document.class
What else can I do? Is there any way to get more information about the
classes available to the particular
Here is the stack trace: (Sorry for the duplicate, Igor -- I forgot
to include the list.)
15/09/08 05:56:43 WARN scheduler.TaskSetManager: Lost task 183.0 in
stage 41.0 (TID 193386, ds-compute2.lumiata.com): java.io.IOException:
com.esotericsoftware.kryo.KryoException: Error constructing
java.lang.ClassNotFoundException: com.i2028.Document.Document
1. so have you checked that jar that you create(fat jar) contains this class?
2. might be there is some stale cache issue...not sure though
On 8 September 2015 at 16:12, Nicholas R. Peterson
wrote:
> Here is
hmm...out of ideas.
can you check in spark ui environment tab that this jar is not somehow
appears 2 times or more...? or more generally - any 2 jars that can contain
this class by any chance
regarding your question about classloader - no idea, probably there is, I
remember stackoverflow has some
another idea - you can add this fat jar explicitly to the classpath of
executors...it's not a solution, but might be it work...
I mean place it somewhere locally on executors and add it to cp with
spark.executor.extraClassPath
On 8 September 2015 at 18:30, Nick Peterson
Yeah... none of the jars listed on the classpath contain this class. The
only jar that does is the fat jar that I'm submitting with spark-submit,
which as mentioned isn't showing up on the classpath anywhere.
-- Nick
On Tue, Sep 8, 2015 at 8:26 AM Igor Berman wrote:
>
Yes, putting the jar on each node and adding it manually to the executor
classpath does it. So, it seems that's where the issue lies.
I'll do some experimenting and see if I can narrow down the problem; but,
for now, at least I can run my job!
Thanks for your help.
On Tue, Sep 8, 2015 at 8:40
The problem which we have now is skew data (2360 tasks done in 5 min, 3
tasks in 40 min and 1 task in 2 hours)
Some people from the team worry that the executor which runs the longest
task can be killed by YARN (because executor might be unresponsive because
of GC or it might occupy more memory
Hi Alex,
If they're both configured correctly, there's no reason that Spark
Standalone should provide performance or memory improvement over Spark on
YARN.
-Sandy
On Fri, Sep 4, 2015 at 1:24 PM, Alexander Pivovarov
wrote:
> Hi Everyone
>
> We are trying the latest aws
Hi Sandy
Thank you for your reply
Currently we use r3.2xlarge boxes (vCPU: 8, Mem: 61 GiB)
with emr setting for Spark "maximizeResourceAllocation": "true"
It is automatically converted to Spark settings
spark.executor.memory47924M
spark.yarn.executor.memoryOverhead 5324
we also set
Yes, you can set the SPARK_LOCAL_DIR in the spark-env.sh or spark.local.dir
in the spark-defaults.conf file, then it would use this location for the
shuffle writes etc.
Thanks
Best Regards
On Wed, Aug 26, 2015 at 6:56 PM, michael.engl...@nomura.com wrote:
Hi,
I am having issues with /tmp
Hi,
I have looked at the UI scheduler tab and it appears my new user was
allocated less cores than my other user, is there any way i can avoid this
happening?
Thanks,
Jem
On Sat, Aug 8, 2015 at 8:32 PM Shushant Arora shushantaror...@gmail.com
wrote:
which is the scheduler on your cluster.
Hi Jem,
Do they fail with any particular exception? Does YARN just never end up
giving them resources? Does an application master start? If so, what are
in its logs? If not, anything suspicious in the YARN ResourceManager logs?
-Sandy
On Fri, Aug 7, 2015 at 1:48 AM, Jem Tucker
Hi Sandy,
The application doesn't fail, it gets accepted by yarn but the application
master never starts and the application state never changes to running. I
have checked in the resource manager and node manager logs and nothing
jumps out.
Thanks
Jem
On Sat, 8 Aug 2015 at 09:20, Sandy Ryza
Hi dustin,
Yes there are enough resources available, the same application run with a
different user works fine so I think it is something to do with permissions
but I can't work out where.
Thanks ,
Jem
On Sat, 8 Aug 2015 at 17:35, Dustin Cote dc...@cloudera.com wrote:
Hi Jem,
In the top of
which is the scheduler on your cluster. Just check on RM UI scheduler tab
and see your user and max limit of vcores for that user , is currently
other applications of that user have occupies till max vcores of this user
then that could be the reason of not allocating vcores to this user but for
it. If you met similar problem, you
could increase this configuration “yarn.nodemanager.vmem-pmem-ratio”.
Thanks
Jerry
*From:* Jeff Zhang [mailto:zjf...@gmail.com]
*Sent:* Thursday, July 30, 2015 4:36 PM
*To:* Jeetendra Gangele
*Cc:* user
*Subject:* Re: Spark on YARN
15/07/30 12:13:35
I can't see the application logs here. All the logs are going into stderr.
can anybody help here?
On 30 July 2015 at 12:21, Jeetendra Gangele gangele...@gmail.com wrote:
I am running below command this is default spark PI program but this is
not running all the log are going in stderr but at
15/07/30 12:13:35 ERROR yarn.ApplicationMaster: RECEIVED SIGNAL 15:
SIGTERM
AM is killed somehow, may due to preemption. Does it always happen ?
Resource manager log would be helpful.
On Thu, Jul 30, 2015 at 4:17 PM, Jeetendra Gangele gangele...@gmail.com
wrote:
I can't see the application
Gangele
Cc: user
Subject: Re: Spark on YARN
15/07/30 12:13:35 ERROR yarn.ApplicationMaster: RECEIVED SIGNAL 15: SIGTERM
AM is killed somehow, may due to preemption. Does it always happen ? Resource
manager log would be helpful.
On Thu, Jul 30, 2015 at 4:17 PM, Jeetendra Gangele
gangele
On Tue, Jul 14, 2015 at 9:57 AM, Shushant Arora shushantaror...@gmail.com
wrote:
When I specify --executor-cores 4 it fails to start the application.
When I give --executor-cores as 4 , it works fine.
Do you have any NM that advertises more than 4 available cores?
Also, it's always worth it
Ok thanks a lot!
few more doubts :
What happens in a streaming application say with
spark-submit --class classname --num-executors 10 --executor-cores 4
--master masteradd jarname
Will it allocate 10 containers throughout the life of streaming application
on same nodes until any node failure
On Tue, Jul 14, 2015 at 11:13 AM, Shushant Arora shushantaror...@gmail.com
wrote:
spark-submit --class classname --num-executors 10 --executor-cores 4
--master masteradd jarname
Will it allocate 10 containers throughout the life of streaming
application on same nodes until any node failure
On Tue, Jul 14, 2015 at 12:03 PM, Shushant Arora shushantaror...@gmail.com
wrote:
Can a container have multiple JVMs running in YARN?
Yes and no. A container runs a single command, but that process can start
other processes, and those also count towards the resource usage of the
container
Can a container have multiple JVMs running in YARN?
I am comparing Hadoop Mapreduce running on yarn vs spark running on yarn
here :
1.Is the difference is in Hadoop Mapreduce job - say I specify 20 reducers
and my job uses 10 map tasks then, it need total 30 containers or 30 vcores
? I guess 30
On Tue, Jul 14, 2015 at 10:40 AM, Shushant Arora shushantaror...@gmail.com
wrote:
My understanding was --executor-cores(5 here) are maximum concurrent
tasks possible in an executor and --num-executors (10 here)are no of
executors or containers demanded by Application master/Spark driver
Is yarn.scheduler.maximum-allocation-vcores the setting for max vcores per
container?
Whats the setting for max limit of --num-executors ?
On Tue, Jul 14, 2015 at 11:18 PM, Marcelo Vanzin van...@cloudera.com
wrote:
On Tue, Jul 14, 2015 at 10:40 AM, Shushant Arora
shushantaror...@gmail.com
Shushant :
Please also see 'Debugging your Application' section of
https://spark.apache.org/docs/latest/running-on-yarn.html
On Tue, Jul 14, 2015 at 10:48 AM, Marcelo Vanzin van...@cloudera.com
wrote:
On Tue, Jul 14, 2015 at 10:40 AM, Shushant Arora
shushantaror...@gmail.com wrote:
My
On Tue, Jul 14, 2015 at 10:55 AM, Shushant Arora shushantaror...@gmail.com
wrote:
Is yarn.scheduler.maximum-allocation-vcores the setting for max vcores per
container?
I don't remember YARN config names by heart, but that sounds promising. I'd
look at the YARN documentation for details.
got the below exception in logs:
org.apache.hadoop.ipc.RemoteException(org.apache.hadoop.yarn.exceptions.InvalidResourceRequestException):
Invalid resource request, requested virtual cores 0, or requested virtual
cores max configured, requestedVirtualCores=5, maxVirtualCores=4
at
Hi Ashish,
For Spark on YARN, you actually only need the Spark files on one machine -
the submission client. This machine could even live outside of the cluster.
Then all you need to do is point YARN_CONF_DIR to the directory containing
your hadoop configuration files (e.g. yarn-site.xml) on that
Hi,
We switched from ParallelGC to CMS, and the symptom is gone.
On Thu, Jun 4, 2015 at 3:37 PM, Ji ZHANG zhangj...@gmail.com wrote:
Hi,
I set spark.shuffle.io.preferDirectBufs to false in SparkConf and this
setting can be seen in web ui's environment tab. But, it still eats memory,
i.e.
btw, user listt will be a better place for this thread.
On Thu, Jun 18, 2015 at 8:19 AM, Yin Huai yh...@databricks.com wrote:
Is it the full stack trace?
On Thu, Jun 18, 2015 at 6:39 AM, Sea 261810...@qq.com wrote:
Hi, all:
I want to run spark sql on yarn(yarn-client), but ... I already
Glad to hear that. :)
On Thu, Jun 18, 2015 at 6:25 AM, Ji ZHANG zhangj...@gmail.com wrote:
Hi,
We switched from ParallelGC to CMS, and the symptom is gone.
On Thu, Jun 4, 2015 at 3:37 PM, Ji ZHANG zhangj...@gmail.com wrote:
Hi,
I set spark.shuffle.io.preferDirectBufs to false in
Hi,
I set spark.shuffle.io.preferDirectBufs to false in SparkConf and this
setting can be seen in web ui's environment tab. But, it still eats memory,
i.e. -Xmx set to 512M but RES grows to 1.5G in half a day.
On Wed, Jun 3, 2015 at 12:02 PM, Shixiong Zhu zsxw...@gmail.com wrote:
Could you
Just testing with Spark 1.3, it looks like it sets the proxy correctly to
be the YARN RM host (0101)
15/06/03 10:34:19 INFO yarn.ApplicationMaster: Registered signal handlers
for [TERM, HUP, INT]
15/06/03 10:34:20 INFO yarn.ApplicationMaster: ApplicationAttemptId:
That code hasn't changed at all between 1.3 and 1.4; it also has been
working fine for me.
Are you sure you're using exactly the same Hadoop libraries (since you're
building with -Phadoop-provided) and Hadoop configuration in both cases?
On Tue, Jun 2, 2015 at 5:29 PM, Night Wolf
Hi,
Thanks for you information. I'll give spark1.4 a try when it's released.
On Wed, Jun 3, 2015 at 11:31 AM, Tathagata Das t...@databricks.com wrote:
Could you try it out with Spark 1.4 RC3?
Also pinging, Cloudera folks, they may be aware of something.
BTW, the way I have debugged memory
Could you try it out with Spark 1.4 RC3?
Also pinging, Cloudera folks, they may be aware of something.
BTW, the way I have debugged memory leaks in the past is as follows.
Run with a small driver memory, say 1 GB. Periodically (maybe a script),
take snapshots of histogram and also do memory
Hi,
Thanks for you reply. Here's the top 30 entries of jmap -histo:live result:
num #instances #bytes class name
--
1: 40802 145083848 [B
2: 99264 12716112 methodKlass
3: 99264 12291480
Thanks Marcelo - looks like it was my fault. Seems when we deployed the new
version of spark it was picking up the wrong yarn site and setting the
wrong proxy host. All good now!
On Wed, Jun 3, 2015 at 11:01 AM, Marcelo Vanzin van...@cloudera.com wrote:
That code hasn't changed at all between
Hi Zhang,
Could you paste your code in a gist? Not sure what you are doing inside the
code to fill up memory.
Thanks
Best Regards
On Thu, May 28, 2015 at 10:08 AM, Ji ZHANG zhangj...@gmail.com wrote:
Hi,
Yes, I'm using createStream, but the storageLevel param is by default
Hi,
I wrote a simple test job, it only does very basic operations. for example:
val lines = KafkaUtils.createStream(ssc, zkQuorum, group, Map(topic -
1)).map(_._2)
val logs = lines.flatMap { line =
try {
Some(parse(line).extract[Impression])
} catch {
case _:
Can you replace your counting part with this?
logs.filter(_.s_id 0).foreachRDD(rdd = logger.info(rdd.count()))
Thanks
Best Regards
On Thu, May 28, 2015 at 1:02 PM, Ji ZHANG zhangj...@gmail.com wrote:
Hi,
I wrote a simple test job, it only does very basic operations. for example:
Hi,
Unfortunately, they're still growing, both driver and executors.
I run the same job with local mode, everything is fine.
On Thu, May 28, 2015 at 5:26 PM, Akhil Das ak...@sigmoidanalytics.com
wrote:
Can you replace your counting part with this?
logs.filter(_.s_id 0).foreachRDD(rdd =
After submitting the job, if you do a ps aux | grep spark-submit then you
can see all JVM params. Are you using the highlevel consumer (receiver
based) for receiving data from Kafka? In that case if your throughput is
high and the processing delay exceeds batch interval then you will hit this
Hi,
Yes, I'm using createStream, but the storageLevel param is by default
MEMORY_AND_DISK_SER_2. Besides, the driver's memory is also growing. I
don't think Kafka messages will be cached in driver.
On Thu, May 28, 2015 at 12:24 AM, Akhil Das ak...@sigmoidanalytics.com
wrote:
Are you using the
Hi Akhil,
Thanks for your reply. Accoding to the Streaming tab of Web UI, the
Processing Time is around 400ms, and there's no Scheduling Delay, so I
suppose it's not the Kafka messages that eat up the off-heap memory. Or
maybe it is, but how to tell?
I googled about how to check the off-heap
Neither of those two. Instead, the shuffle data is cleaned up when the
stage they are from get GC'ed by the jvm. that is, when you are no longer
holding any references to anything which points to the old stages, and
there is an appropriate gc event.
The data is not cleaned up right after the
Thanks for the quick reply.
I am running the application in YARN client mode.
And I want to run the AM on the same node as RM inorder use the node which
otherwise would run AM.
How can I get AM run on the same node as RM?
On Tue, Mar 10, 2015 at 3:49 PM, Sean Owen so...@cloudera.com wrote:
I suppose you just provision enough resource to run both on that
node... but it really shouldn't matter. The RM and your AM aren't
communicating heavily.
On Tue, Mar 10, 2015 at 10:23 AM, Harika Matha matha.har...@gmail.com wrote:
Thanks for the quick reply.
I am running the application in
In YARN cluster mode, there is no Spark master, since YARN is your
resource manager. Yes you could force your AM somehow to run on the
same node as the RM, but why -- what do think is faster about that?
On Tue, Mar 10, 2015 at 10:06 AM, Harika matha.har...@gmail.com wrote:
Hi all,
I have Spark
The version I'm using was already pre-built for Hadoop 2.3.
--
View this message in context:
http://apache-spark-user-list.1001560.n3.nabble.com/Spark-on-Yarn-java-lang-IllegalArgumentException-Invalid-rule-tp21382p21485.html
Sent from the Apache Spark User List mailing list archive at
Can you look inside RM log to see if you can find some clue there ?
You can pastebin part of the RM log around the time your job ran ?
What hadoop version are you using ?
Thanks
On Sat, Jan 31, 2015 at 11:24 AM, Koert Kuipers ko...@tresata.com wrote:
i have a simple spark app that i run with
it is CDH 5.3 with the spark that ships with it.
i went through the RM logs line by line, and i found the exit code in there:
container_1422728945460_0001_01_29 Container Transitioned from NEW to
RESERVED
2015-01-31 18:30:49,633 INFO
Here is the same issues:
[1]
http://stackoverflow.com/questions/28186607/java-lang-classcastexception-using-lambda-expressions-in-spark-job-on-remote-ser
[2]
Then your spark is not built for yarn. Try to build with
sbt/sbt -Dhadoop.version=2.3.0 -Pyarn assembly
--
View this message in context:
http://apache-spark-user-list.1001560.n3.nabble.com/Spark-on-Yarn-java-lang-IllegalArgumentException-Invalid-rule-tp21382p21404.html
Sent from the Apache
Thanks, Ted. Kerberos is enabled on the cluster.
I'm new to the world of kerberos, so pease excuse my ignorance here. Do you
know if there are any additional steps I need to take in addition to
setting HADOOP_CONF_DIR? For instance, does hadoop.security.auth_to_local
require any specific setting
Thanks, Siddardha. I did but got the same error. Kerberos is enabled on my
cluster and I may be missing a configuration step somewhere.
--
View this message in context:
Caused by: java.lang.IllegalArgumentException: Invalid rule: L
RULE:[2:$1@$0](.*@XXXCOMPANY.COM http://xxxcompany.com/)s/(.*)@
XXXCOMPANY.COM/$1/L http://xxxcompany.com/$1/L
DEFAULT
Can you put the rule on a single line (not sure whether there is newline or
space between L and DEFAULT) ?
Looks
Update: I deployed a stand-alone spark in localhost then set Master as
spark://localhost:7077 and it met the same issue
Don't know how to solve it.
--
View this message in context:
Actually it does causes builds with SBT 0.13.7 to fail with the error
Conflicting cross-version suffixes. I have raised a defect SPARK-5143 for
this.
On Wed Jan 07 2015 at 23:44:21 Marcelo Vanzin van...@cloudera.com wrote:
This particular case shouldn't cause problems since both of those
This particular case shouldn't cause problems since both of those
libraries are java-only (the scala version appended there is just for
helping the build scripts).
But it does look weird, so it would be nice to fix it.
On Wed, Jan 7, 2015 at 12:25 AM, Aniket Bhatnagar
aniket.bhatna...@gmail.com
/JW1q5vd61V1/Spark-yarn+1.2.0subj=Re+spark+yarn_2+10+1+2+0+artifacts
Cheers
On Dec 28, 2014, at 11:13 PM, Aniket Bhatnagar aniket.bhatna...@gmail.com
wrote:
Hi all
I just realized that spark-yarn artifact hasn't been published for 1.2.0
release. Any particular reason for that? I was using
See this thread:
http://search-hadoop.com/m/JW1q5vd61V1/Spark-yarn+1.2.0subj=Re+spark+yarn_2+10+1+2+0+artifacts
Cheers
On Dec 28, 2014, at 11:13 PM, Aniket Bhatnagar aniket.bhatna...@gmail.com
wrote:
Hi all
I just realized that spark-yarn artifact hasn't been published for 1.2.0
release
Thanks Sandy!
On Mon, Dec 8, 2014 at 23:15 Sandy Ryza sandy.r...@cloudera.com wrote:
Another thing to be aware of is that YARN will round up containers to the
nearest increment of yarn.scheduler.minimum-allocation-mb, which defaults
to 1024.
-Sandy
On Sat, Dec 6, 2014 at 3:48 PM, Denny Lee
Another thing to be aware of is that YARN will round up containers to the
nearest increment of yarn.scheduler.minimum-allocation-mb, which defaults
to 1024.
-Sandy
On Sat, Dec 6, 2014 at 3:48 PM, Denny Lee denny.g@gmail.com wrote:
Got it - thanks!
On Sat, Dec 6, 2014 at 14:56 Arun Ahuja
Hi Denny,
This is due the spark.yarn.memoryOverhead parameter, depending on what
version of Spark you are on the default of this may differ, but it should
be the larger of 1024mb per executor or .07 * executorMemory.
When you set executor memory, the yarn resource request is executorMemory +
Got it - thanks!
On Sat, Dec 6, 2014 at 14:56 Arun Ahuja aahuj...@gmail.com wrote:
Hi Denny,
This is due the spark.yarn.memoryOverhead parameter, depending on what
version of Spark you are on the default of this may differ, but it should
be the larger of 1024mb per executor or .07 *
Hi all!
Thanks for answering!
@Sean, I tried to run with 30 executor-cores , and 1 machine still without
processing.
@Vanzin, I checked RM's web UI, and all nodes were detecteds and RUNNING.
The interesting fact is that available
memory and available core of 1 node was different of other 2, with
I think your config may be the issue then. It sounds like 1 server is
configured in a different YARN group that states it has way less
resource than it does.
On Wed, Nov 19, 2014 at 5:27 PM, Alan Prando a...@scanboo.com.br wrote:
Hi all!
Thanks for answering!
@Sean, I tried to run with 30
Not sure how to solve this, but spotted these lines in the logs:
14/11/18 14:28:23 INFO YarnAllocationHandler: Container marked as
*failed*: container_1415961020140_0325_01_02
14/11/18 14:28:38 INFO YarnAllocationHandler: Container marked as
*failed*: container_1415961020140_0325_01_03
I run my Spark on YARN jobs as:
HADOOP_CONF_DIR=/etc/hadoop/conf/ /app/data/v606014/dist/bin/spark-submit
--master yarn --jars test-job.jar --executor-cores 4 --num-executors 10
--executor-memory 16g --driver-memory 4g --class TestClass test.jar
It uses HADOOP_CONF_DIR to schedule executors and
Can you check in your RM's web UI how much of each resource does Yarn
think you have available? You can also check that in the Yarn
configuration directly.
Perhaps it's not configured to use all of the available resources. (If
it was set up with Cloudera Manager, CM will reserve some room for
Hey Alan,
Spark's application master will take up 1 core on one of the nodes on the
cluster. This means that that node will only have 31 cores remaining, not
enough to fit your third executor.
-Sandy
On Tue, Nov 18, 2014 at 10:03 AM, Alan Prando a...@scanboo.com.br wrote:
Hi Folks!
I'm
My guess is you're asking for all cores of all machines but the driver
needs at least one core, so one executor is unable to find a machine to fit
on.
On Nov 18, 2014 7:04 PM, Alan Prando a...@scanboo.com.br wrote:
Hi Folks!
I'm running Spark on YARN cluster installed with Cloudera Manager
So it seems that this problem was related to
http://apache-spark-developers-list.1001551.n3.nabble.com/Lost-executor-on-YARN-ALS-iterations-td7916.html
and increasing the executor memory worked for me.
__
Hi,
I am getting
I have tried it out to merge the file to one, Spark is now working with RAM as
I've expected.
Unfortunately after doing this there appears another problem. Now Spark running
on YARN is scheduling all the work only to one worker node as a one big job. Is
there some way, how to force Spark and
Ok so the problem was solved, it that the file was gziped and it looks that
Spark does not support direct .gz file distribution to workers.
Thank you very much fro the suggestion to merge the files.
Best regards,
Jan
__
I have
Could you please give me an example or send me a link of how to use Hadoop
CombinedFileInputFormat? It sound very interesting to me and it would probably
save me several hours of my pipeline computation. Merging of the files is
currently the bottleneck in my system.
On Sun, Nov 2, 2014 at 1:35 AM, jan.zi...@centrum.cz wrote:
Hi,
I am using Spark on Yarn, particularly Spark in Python. I am trying to run:
myrdd = sc.textFile(s3n://mybucket/files/*/*/*.json)
How many files do you have? and the average size of each file?
myrdd.getNumPartitions()
I have 3 datasets in all the datasets the average file size is 10-12Kb.
I am able to run my code on the dataset with 70K files, but I am not able to
run it on datasets with 1.1M and 3.8M files.
__
On Sun, Nov 2, 2014 at 1:35 AM,
We have narrowed this hanging issue down to the calliope package
that we used to create RDD from reading cassandra table.
The calliope native RDD interface seems hanging and I have decided to switch
to the calliope cql3 RDD interface.
--
View this message in context:
It may also cause a problem when running in the yarn-client mode. If
--driver-memory is large, Yarn has to allocate a lot of memory to the AM
container, but AM doesn't really need the memory.
Boduo
--
View this message in context:
: Andrew Or and...@databricks.commailto:and...@databricks.com
Date: Wednesday, October 8, 2014 3:25 PM
To: Greg greg.h...@rackspace.commailto:greg.h...@rackspace.com
Cc: user@spark.apache.orgmailto:user@spark.apache.org
user@spark.apache.orgmailto:user@spark.apache.org
Subject: Re: Spark on YARN driver
@spark.apache.org user@spark.apache.org
Subject: Re: Spark on YARN driver memory allocation bug?
Hi Greg,
It does seem like a bug. What is the particular exception message that
you see?
Andrew
2014-10-08 12:12 GMT-07:00 Greg Hill greg.h...@rackspace.com:
So, I think this is a bug, but I wanted
101 - 200 of 239 matches
Mail list logo