Re: Spark-sql versus Impala versus Hive
Interesting. What where the Hive settings? Specifically it would be useful to know if this was Hive on Tez. - Steve From: Sanjay Subramanian Reply-To: Sanjay Subramanian Date: Thursday, June 18, 2015 at 11:08 To: user@spark.apache.orgmailto:user@spark.apache.org Subject: Spark-sql versus Impala versus Hive I just published results of my findings here https://bigdatalatte.wordpress.com/2015/06/18/spark-sql-versus-impala-versus-hive/
Re: Pairwise Processing of a List
Not combinations, linear distances, e.g., given: List[ (x1,y1), (x2,y2), (x3,y3) ], compute the sum of: distance (x1,y2) and (x2,y2) and distance (x2,y2) and (x3,y3) Imagine that the list of coordinate point comes from a GPS and describes a trip. - Steve From: Joseph Lust jl...@mc10inc.commailto:jl...@mc10inc.com Date: Sunday, January 25, 2015 at 17:17 To: Steve Nunez snu...@hortonworks.commailto:snu...@hortonworks.com, user@spark.apache.orgmailto:user@spark.apache.org user@spark.apache.orgmailto:user@spark.apache.org Subject: Re: Pairwise Processing of a List So you've got a point A and you want the sum of distances between it and all other points? Or am I misunderstanding you? // target point, can be Broadcast global sent to all workers val tarPt = (10,20) val pts = Seq((2,2),(3,3),(2,3),(10,2)) val rdd= sc.parallelize(pts) rdd.map( pt = Math.sqrt( Math.pow(tarPt._1 - pt._1,2) + Math.pow(tarPt._2 - pt._2,2)) ).reduce( (d1,d2) = d1+d2) -Joe From: Steve Nunez snu...@hortonworks.commailto:snu...@hortonworks.com Date: Sunday, January 25, 2015 at 7:32 PM To: user@spark.apache.orgmailto:user@spark.apache.org user@spark.apache.orgmailto:user@spark.apache.org Subject: Pairwise Processing of a List Spark Experts, I've got a list of points: List[(Float, Float)]) that represent (x,y) coordinate pairs and need to sum the distance. It's easy enough to compute the distance: case class Point(x: Float, y: Float) { def distance(other: Point): Float = sqrt(pow(x - other.x, 2) + pow(y - other.y, 2)).toFloat } (in this case I create a 'Point' class, but the maths are the same). What I can't figure out is the 'right' way to sum distances between all the points. I can make this work by traversing the list with a for loop and using indices, but this doesn't seem right. Anyone know a clever way to process List[(Float, Float)]) in a pairwise fashion? Regards, - Steve CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
Pairwise Processing of a List
Spark Experts, I've got a list of points: List[(Float, Float)]) that represent (x,y) coordinate pairs and need to sum the distance. It's easy enough to compute the distance: case class Point(x: Float, y: Float) { def distance(other: Point): Float = sqrt(pow(x - other.x, 2) + pow(y - other.y, 2)).toFloat } (in this case I create a 'Point' class, but the maths are the same). What I can't figure out is the 'right' way to sum distances between all the points. I can make this work by traversing the list with a for loop and using indices, but this doesn't seem right. Anyone know a clever way to process List[(Float, Float)]) in a pairwise fashion? Regards, - Steve
Directory / File Reading Patterns
Hello Users, I've got a real-world use case that seems common enough that its pattern would be documented somewhere, but I can't find any references to a simple solution. The challenge is that data is getting dumped into a directory structure, and that directory structure itself contains features that I need in my model. For example: bank_code Trader Day-1.csv Day-2.csv ... Each CVS file contains a list of all the trades made by that individual each day. The problem is that the bank trader should be part of the feature set. I.e. We need the RDD to look like: (bank, trader, day, list-of-trades) Anyone got any elegant solutions for doing this? Cheers, - SteveN
Re: Breaking the previous large-scale sort record with Spark
Great stuff. Wonderful to see such progress in so short a time. How about some links to code and instructions so that these benchmarks can be reproduced? Regards, - Steve From: Debasish Das debasish.da...@gmail.com Date: Friday, October 10, 2014 at 8:17 To: Matei Zaharia matei.zaha...@gmail.com Cc: user user@spark.apache.org, dev d...@spark.apache.org Subject: Re: Breaking the previous large-scale sort record with Spark Awesome news Matei ! Congratulations to the databricks team and all the community members... On Fri, Oct 10, 2014 at 7:54 AM, Matei Zaharia matei.zaha...@gmail.com wrote: Hi folks, I interrupt your regularly scheduled user / dev list to bring you some pretty cool news for the project, which is that we've been able to use Spark to break MapReduce's 100 TB and 1 PB sort records, sorting data 3x faster on 10x fewer nodes. There's a detailed writeup at http://databricks.com/blog/2014/10/10/spark-breaks-previous-large-scale-sort- record.html. Summary: while Hadoop MapReduce held last year's 100 TB world record by sorting 100 TB in 72 minutes on 2100 nodes, we sorted it in 23 minutes on 206 nodes; and we also scaled up to sort 1 PB in 234 minutes. I want to thank Reynold Xin for leading this effort over the past few weeks, along with Parviz Deyhim, Xiangrui Meng, Aaron Davidson and Ali Ghodsi. In addition, we'd really like to thank Amazon's EC2 team for providing the machines to make this possible. Finally, this result would of course not be possible without the many many other contributions, testing and feature requests from throughout the community. For an engine to scale from these multi-hour petabyte batch jobs down to 100-millisecond streaming and interactive queries is quite uncommon, and it's thanks to all of you folks that we are able to make this happen. Matei - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
FW: Reference Accounts Large Node Deployments
Anyone? No customers using streaming at scale? From: Steve Nunez snu...@hortonworks.com Date: Wednesday, August 27, 2014 at 9:08 To: user@spark.apache.org user@spark.apache.org Subject: Reference Accounts Large Node Deployments All, Does anyone have specific references to customers, use cases and large-scale deployments of Spark Streaming? By OElarge scale¹ I mean both through-put and number of nodes. I¹m attempting an objective comparison of Streaming and Storm and while this data is known for Storm, there appears to be little for Spark Streaming. If you know of any such deployments, please post them here because I am sure I¹m not the only one wondering about this. If customer confidentially prevents mentioning them by name, consider identifying them by industry, e.g. OEtelco doing X with streaming using Y nodes¹. Any information at all will be welcome. I¹ll feed back a summary and/or update a wiki page once I collate the information. Cheers, - Steve -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
Reference Accounts Large Node Deployments
All, Does anyone have specific references to customers, use cases and large-scale deployments of Spark Streaming? By OElarge scale¹ I mean both through-put and number of nodes. I¹m attempting an objective comparison of Streaming and Storm and while this data is known for Storm, there appears to be little for Spark Streaming. If you know of any such deployments, please post them here because I am sure I¹m not the only one wondering about this. If customer confidentially prevents mentioning them by name, consider identifying them by industry, e.g. OEtelco doing X with streaming using Y nodes¹. Any information at all will be welcome. I¹ll feed back a summary and/or update a wiki page once I collate the information. Cheers, - Steve -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
Re: Issues with HDP 2.4.0.2.1.3.0-563
I don’t think there is an hwx profile, but there probably should be. - Steve From: Patrick Wendell pwend...@gmail.com Date: Monday, August 4, 2014 at 10:08 To: Ron's Yahoo! zlgonza...@yahoo.com Cc: Ron's Yahoo! zlgonza...@yahoo.com.invalid, Steve Nunez snu...@hortonworks.com, user@spark.apache.org, d...@spark.apache.org d...@spark.apache.org Subject: Re: Issues with HDP 2.4.0.2.1.3.0-563 Ah I see, yeah you might need to set hadoop.version and yarn.version. I thought he profile set this automatically. On Mon, Aug 4, 2014 at 10:02 AM, Ron's Yahoo! zlgonza...@yahoo.com wrote: I meant yarn and hadoop defaulted to 1.0.4 so the yarn build fails since 1.0.4 doesn’t exist for yarn... Thanks, Ron On Aug 4, 2014, at 10:01 AM, Ron's Yahoo! zlgonza...@yahoo.com wrote: That failed since it defaulted the versions for yarn and hadoop I’ll give it a try with just 2.4.0 for both yarn and hadoop… Thanks, Ron On Aug 4, 2014, at 9:44 AM, Patrick Wendell pwend...@gmail.com wrote: Can you try building without any of the special `hadoop.version` flags and just building only with -Phadoop-2.4? In the past users have reported issues trying to build random spot versions... I think HW is supposed to be compatible with the normal 2.4.0 build. On Mon, Aug 4, 2014 at 8:35 AM, Ron's Yahoo! zlgonza...@yahoo.com.invalid wrote: Thanks, I ensured that $SPARK_HOME/pom.xml had the HDP repository under the repositories element. I also confirmed that if the build couldn’t find the version, it would fail fast so it seems as if it’s able to get the versions it needs to build the distribution. I ran the following (generated from make-distribution.sh), but it did not address the problem, while building with an older version (2.4.0.2.1.2.0-402) worked. Any other thing I can try? mvn clean package -Phadoop-2.4 -Phive -Pyarn -Dyarn.version=2.4.0.2.1.2.0-563 -Dhadoop.version=2.4.0.2.1.3.0-563 -DskipTests Thanks, Ron On Aug 4, 2014, at 7:13 AM, Steve Nunez snu...@hortonworks.com wrote: Provided you¹ve got the HWX repo in your pom.xml, you can build with this line: mvn -Pyarn -Phive -Phadoop-2.4 -Dhadoop.version=2.4.0.2.1.1.0-385 -DskipTests clean package I haven¹t tried building a distro, but it should be similar. - SteveN On 8/4/14, 1:25, Sean Owen so...@cloudera.com wrote: For any Hadoop 2.4 distro, yes, set hadoop.version but also set -Phadoop-2.4. http://spark.apache.org/docs/latest/building-with-maven.html On Mon, Aug 4, 2014 at 9:15 AM, Patrick Wendell pwend...@gmail.com wrote: For hortonworks, I believe it should work to just link against the corresponding upstream version. I.e. just set the Hadoop version to 2.4.0 Does that work? - Patrick On Mon, Aug 4, 2014 at 12:13 AM, Ron's Yahoo! zlgonza...@yahoo.com.invalid wrote: Hi, Not sure whose issue this is, but if I run make-distribution using HDP 2.4.0.2.1.3.0-563 as the hadoop version (replacing it in make-distribution.sh), I get a strange error with the exception below. If I use a slightly older version of HDP (2.4.0.2.1.2.0-402) with make-distribution, using the generated assembly all works fine for me. Either 1.0.0 or 1.0.1 will work fine. Should I file a JIRA or is this a known issue? Thanks, Ron Exception in thread main org.apache.spark.SparkException: Job aborted due to stage failure: Task 0.0:0 failed 1 times, most recent failure: Exception failure in TID 0 on host localhost: java.lang.IncompatibleClassChangeError: Found interface org.apache.hadoop.mapreduce.TaskAttemptContext, but class was expected org.apache.avro.mapreduce.AvroKeyInputFormat.createRecordReader(AvroKeyI nputFormat.java:47) org.apache.spark.rdd.NewHadoopRDD$$anon$1.init(NewHadoopRDD.scala:111) org.apache.spark.rdd.NewHadoopRDD.compute(NewHadoopRDD.scala:99) org.apache.spark.rdd.NewHadoopRDD.compute(NewHadoopRDD.scala:61) org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:262) org.apache.spark.rdd.RDD.iterator(RDD.scala:229) org.apache.spark.rdd.MappedRDD.compute(MappedRDD.scala:31) org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:262) org.apache.spark.CacheManager.getOrCompute(CacheManager.scala:77) org.apache.spark.rdd.RDD.iterator(RDD.scala:227) org.apache.spark.rdd.MappedRDD.compute(MappedRDD.scala:31) org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:262) org.apache.spark.rdd.RDD.iterator(RDD.scala:229) org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:111) org.apache.spark.scheduler.Task.run(Task.scala:51) org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:187) java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.jav a:1145) java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.ja va:615) java.lang.Thread.run(Thread.java:745
Re: Issues with HDP 2.4.0.2.1.3.0-563
Hmm. Fair enough. I hadn¹t given that answer much thought and on reflection think you¹re right in that a profile would just be a bad hack. On 8/4/14, 10:35, Sean Owen so...@cloudera.com wrote: What would such a profile do though? In general building for a specific vendor version means setting hadoop.verison and/or yarn.version. Any hard-coded value is unlikely to match what a particular user needs. Setting protobuf versions and so on is already done by the generic profiles. In a similar vein, I am not clear on why there's a mapr profile in the build. Its versions are about to be out of date and won't work with upcoming Hbase changes for example. (Elsewhere in the build I think it wouldn't hurt to clear out cloudera-specific profiles and releases too -- they're not in the pom but are in the distribution script. It's the vendor's problem.) This isn't any argument about being purist but just that I am not sure these are things that the project can meaningfully bother with. It makes sense to set vendor repos in the pom for convenience, and makes sense to run smoke tests in Jenkins against particular versions. $0.02 Sean On Mon, Aug 4, 2014 at 6:21 PM, Steve Nunez snu...@hortonworks.com wrote: I don¹t think there is an hwx profile, but there probably should be. -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
MovieLensALS - Scala Pattern Magic
Can one of the Scala experts please explain this bit of pattern magic from the Spark ML tutorial: _._2.user ? As near as I can tell, this is applying the _2 function to the wildcard, and then applying the Œuser¹ function to that. In a similar way the Œproduct¹ function is applied in the next line, yet these functions don¹t seem to exist anywhere in the project, nor are they used anywhere else in the code. It almost makes sense, but not quite. Code below: val ratings = sc.textFile(new File(movieLensHomeDir, ratings.dat).toString).map { line = val fields = line.split(::) // format: (timestamp % 10, Rating(userId, movieId, rating)) (fields(3).toLong % 10, Rating(fields(0).toInt, fields(1).toInt, fields(2).toDouble)) } Š val numRatings = ratings.count val numUsers = ratings.map(_._2.user).distinct.count val numMovies = ratings.map(_._2.product).distinct.count Cheers, - Steve Nunez -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
Emacs Setup Anyone?
Anyone out there have a good configuration for emacs? Scala-mode sort of works, but I¹d love to see a fully-supported spark-mode with an inferior shell. Searching didn¹t turn up much of anything. Any emacs users out there? What setup are you using? Cheers, - SteveN -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
Re: Cluster submit mode - only supported on Yarn?
I¹m also in early stages of setting up long running Spark jobs. Easiest way I¹ve found is to set up a cluster and submit the job via YARN. Then I can come back and check in on progress when I need to. Seems the trick is tuning the queue priority and YARN preemption to get the job to run in a reasonable amount of time without disrupting the other jobs. - SteveN From: Chris Schneider ch...@christopher-schneider.com Reply-To: user@spark.apache.org Date: Wednesday, July 23, 2014 at 7:39 To: user@spark.apache.org Subject: Cluster submit mode - only supported on Yarn? We are starting to use Spark, but we don't have any existing infrastructure related to big-data, so we decided to setup the standalone cluster, rather than mess around with Yarn or Mesos. But it appears like the driver program has to stay up on the client for the full duration of the job (client mode). What is the simplest way to setup cluster submission mode, to allow our client boxes to submit jobs and then move on with the other work they need to do without keeping a potentially long running java process up? Thanks, Chris -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.