AW: Operating on Serialized Data

2015-02-25 Thread Kirschnick, Johannes
Hi Max, thanks for the detailed answer- That was exactly what I have been looking for. I switched the serialization from kryo to use the Value interface instead, keeping everything constant basically halved the execution time - nice. One note - iterating over the array in serial fashion works, b

Re: Stale Synchronous Parallel iterations in Flink

2015-02-25 Thread Martin Neumann
Hej, Very interesting discussion. I hadn't heard of the SSP model before, looks like something I want to look into. I wonder if any of the algorithms that would work in that model would not work in an asynchronous model. Since asynchronous is basically a SSP model with infinite slack. Iterative co

Understanding and Contributing to Flink

2015-02-25 Thread amit pal
Hi Guys, I have been reading the Flink documentation and the mail list discussions since last two days, but haven't completely got hold of what all is there in Flink. My next step is to start with the examples present on github. Can someone give me a gist of what Flink is all about and some deta

Re: Understanding and Contributing to Flink

2015-02-25 Thread Robert Metzger
Hey, since you've already read the documentation, I can recommend checking out some slides about Flink on Slideshare as well. Here is our "How to Contribute" guide: http://flink.apache.org/how-to-contribute.html Best, Robert On Wed, Feb 25, 2015 at 11:09 AM, amit pal wrote: > Hi Guys, > > I h

Re: Understanding and Contributing to Flink

2015-02-25 Thread Max Michels
Hi Amit, Nice to hear you're interested in Flink. The first thing you could do is run the provided examples. Next, you could try to implement a simple Flink job yourself. Besides the "How to contribute" guide, take a look at the material page [1], the Flink blog [2], and the Data Artisans blog [3]

Re: Understanding and Contributing to Flink

2015-02-25 Thread amit pal
Hey all, I will start with some starter JIRA tasks, along with building some application of Flink, particularly in the domain of NLP/text processing. Any good NLT idea to implement over Flink? I am thinking on picking this JIRA one to start on the

Re: Understanding and Contributing to Flink

2015-02-25 Thread Max Michels
Hi Amit, Extensions of Flink should go in the flink-contrib project. Try implementing a simple algorithm for your first task. If you extend it and it becomes useful for other people besides you, make a pull request via GitHub, so other people can check it out. Your general understanding if the Fl

Log4j configuration JUnit vs. Scalatest in IntelliJ

2015-02-25 Thread Ufuk Celebi
Hey Flinksters and IntelliJers, ;-) the tests resources directory of each Maven module contains a log4j-test.properties files, which gets picked via the classpath by JUnit tests, but not Scalatest. Instead Scalatest picks up log4j.properties, but JUnit doesn't. It works when I specify the file

Re: k-means example behavior

2015-02-25 Thread Alexander Alexandrov
Apache's commons-math implementation offers various strategies for handling this scenarios: http://commons.apache.org/proper/commons-math/jacoco/org.apache.commons.math3.stat.clustering/KMeansPlusPlusClusterer.java.html (take a look at the EmptyClusterStrategy enum options) 2015-02-24 23:28 GMT+

Re: Could not build up connection to JobManager

2015-02-25 Thread Dulaj Viduranga
Hi, Sorry for the delay to reply on this issue. the jobmanager.rpc.address is set to “localhost” already in conf.yaml. This can’t be an issue because the job manager web interface works fine which also runs on localhost bin/flink run doesn’t seem to work either. Let me send you my command and

Questions about flink-streaming-examples

2015-02-25 Thread Matthias J. Sax
Hi, I tried to build flink-streaming-examples, but got an compilation error. If I build flink-streaming (that implicitly builds flink-streaming-examples), it works. I tried it on a fresh clone of flink/master using "mvn clean compile test-compile" Can anybody explain this behavior? -Matthias

[jira] [Created] (FLINK-1610) Java docs do not build

2015-02-25 Thread Max Michels (JIRA)
Max Michels created FLINK-1610: -- Summary: Java docs do not build Key: FLINK-1610 URL: https://issues.apache.org/jira/browse/FLINK-1610 Project: Flink Issue Type: Bug Components: Build

Re: Log4j configuration JUnit vs. Scalatest in IntelliJ

2015-02-25 Thread Till Rohrmann
The reason for this behaviour is the following: The log4j-test.properties is not a standard log4j properties file. It is only used if it is explicitly given to the executing JVM by -Dlog4j.configuration. The parent pom defines for the surefire plugin a corresponding argument line with this option.

Re: Questions about flink-streaming-examples

2015-02-25 Thread Max Michels
Hi Matthias, Did you really pull from the latest master? I just tried to compile flink-streaming-examples using "mvn clean compile test-compile" and it worked. Best regards, Max On Wed, Feb 25, 2015 at 4:13 PM, Matthias J. Sax wrote: > Hi, > > I tried to build flink-streaming-examples, but got

Re: Questions about flink-streaming-examples

2015-02-25 Thread Till Rohrmann
Try to do first a mvn install on flink-parent. On Wed, Feb 25, 2015 at 4:54 PM, Max Michels wrote: > Hi Matthias, > > Did you really pull from the latest master? I just tried to compile > flink-streaming-examples using "mvn clean compile test-compile" and it > worked. > > Best regards, > Max > >

Re: Questions about flink-streaming-examples

2015-02-25 Thread Dulaj Viduranga
Yes. Builds fine here too.. Seems like some linking errors.. > On Feb 25, 2015, at 9:24 PM, Max Michels wrote: > > Hi Matthias, > > Did you really pull from the latest master? I just tried to compile > flink-streaming-examples using "mvn clean compile test-compile" and it > worked. > > Best re

Re: Questions about flink-streaming-examples

2015-02-25 Thread Stephan Ewen
Hi Matthias! Can you try "mvn clean package"? It package is generally preferable to compile, in my opinion. There may be a dependency through a test jar that is not properly handled when you do not execute the package goal. Stephan Am 25.02.2015 16:54 schrieb "Max Michels" : > Hi Matthias, > >

Re: Questions about flink-streaming-examples

2015-02-25 Thread Matthias J. Sax
Thanks! Even if I am not a building and/or maven expert, it seems to me that the dependencies are not configures correctly... No clue how to fix it. But it might be a good idea to have a look into it IMHO. -Matthias On 02/25/2015 05:07 PM, Stephan Ewen wrote: > Hi Matthias! > > Can you try "mv

Re: Questions about flink-streaming-examples

2015-02-25 Thread Dulaj Viduranga
Hi, Were you able to solve this? It seams the examples depend on "flink-streaming-core”. I think you have to build it before. > On Feb 25, 2015, at 10:04 PM, Matthias J. Sax > wrote: > > Thanks! > > Even if I am not a building and/or maven expert, it seems to me that the > dependencies are no

[jira] [Created] (FLINK-1611) Rename classes and packages with Nephele

2015-02-25 Thread Henry Saputra (JIRA)
Henry Saputra created FLINK-1611: Summary: Rename classes and packages with Nephele Key: FLINK-1611 URL: https://issues.apache.org/jira/browse/FLINK-1611 Project: Flink Issue Type: Improvemen

Re: Could not build up connection to JobManager

2015-02-25 Thread Stephan Ewen
Okay, the problem seems to be that even though both the client and the jobmanager use "localhost" as the host name, they resolve this to different IP addresses: In one case 127.0.0.1 in the other case 10.216.177.146 Also, the 127.0.0.1 address cannot communicate to 10.216.177.146 apparently. Can

Re: Could not build up connection to JobManager

2015-02-25 Thread Stephan Ewen
Addition: To check whether a port is reachable, I think the easiest thing is to try and connect with a telnet client and see if the connection is refused. On Wed, Feb 25, 2015 at 8:15 PM, Stephan Ewen wrote: > Okay, the problem seems to be that even though both the client and the > jobmanager us

Re: k-means example behavior

2015-02-25 Thread Vasiliki Kalavri
Thanks for the replies guys! @Aljoscha: I get your point, but I would actually expect either an error message or the lonely centroid to move. k-means is supposed to cluster data in k clusters. If you end up with < k, something must have gone wrong.. :s @Alex: very helpful resource, thanks. I will

[jira] [Created] (FLINK-1612) Add guidelines to avoid duplicate class names and more JavaDoc for new addition

2015-02-25 Thread Henry Saputra (JIRA)
Henry Saputra created FLINK-1612: Summary: Add guidelines to avoid duplicate class names and more JavaDoc for new addition Key: FLINK-1612 URL: https://issues.apache.org/jira/browse/FLINK-1612 Project