Re: maven integration in 0.8

Perhaps, but you don't want to version your kafka artifacts as just 2.8.0. The way you handle something like that with maven, is to include the dependency on scala 2.8.0 in the the kafka pom's dependency list. e.g.: org.scala-lang scala-library 2.8.0 But the version of kafka is 0.8.0, not 2.8

Re: maven integration in 0.8

It's actually a rather helpful convention SBT has instituted to deal with bytecode incompatibilities between the different versions of Scala. I wouldn't advise removing it at all - using a 2.8.0-built project in production with Scala 2.9 or higher is courting disaster. -- Alex Zuzin Sent with

Re: maven integration in 0.8

Except that it still has the annoying feature of naming the kafka version after the scala version, which doesn't make sense, e.g.: ~/.ivy2/local/org.apache/kafka_2.8.0/0.8.0-SNAPSHOT/poms ls -l total 24 -rw-r--r-- 1 jbr jbr 3786 May 7 13:43 kafka_2.8.0.pom -rw-r--r-- 1 jbr jbr32 May 7 1

Re: maven integration in 0.8

Alex, Thanks for the tip, this is exactly what I need. Jason On Tue, May 7, 2013 at 11:58 AM, Gray, Alex wrote: > Hi Jason, > > I don't know if this will help you, but I'm pulling the source code of > kafka down to a local machine and calling: > ./sbt update > /sbt -Dsbt.log.noformat=true pub

Kafka Monitoring, 0.7 vs. 0.8 JMX

Hi, We're considering adding Kafka monitoring to SPM (see http://sematext.com/spm/index.html ). We use Kafka ourselves and would like to see our Kafka metrics in SPM along with Hadoop, HBase, Solr, and other metrics we monitor. My questions: * What do people currently use for monitoring Kafka? *

RE: maven integration in 0.8

Hi Jason, I don't know if this will help you, but I'm pulling the source code of kafka down to a local machine and calling: ./sbt update /sbt -Dsbt.log.noformat=true publish-local This will publish the generated pom's and jars to your local ~/.ivy2 directory. Once you have that, you can publish

RE: develop customized producer and consumer in java

Hi Libo, It works for me. I am on 0.8: $ git --no-pager log --max-count=5 --pretty=oneline --decorate e37a4644209de00f265a5361b31ca63cafc895e9 (HEAD) Trivial commit - warn on consumer fetch request errors. 6e640e355632d97e322e462c2bbc379c05d407a4 kafka-880; NoLeaderPartitionSet should be cleare

Re: slow log recovery

That will work Jun (I guess it's not different than the current situation with 0.7.x). (And I still think it should be possible to recover logs in parallel!). Jason On Tue, May 7, 2013 at 7:55 AM, Jun Rao wrote: > If a broker is down, the cluster will be running in under replicated mode, > ie

Re: Kafka C++ Sync Producer on Windows

Pankaj, The non-java client code is maintained separately. You can submit a pull request directly to the client project to make changes. Thanks, Jun On Tue, May 7, 2013 at 7:16 AM, Pankaj Misra wrote: > Hi All, > > I am using Kafka 0.7.2. I was able to run the async example of the C++ > produ

Re: develop customized producer and consumer in java

Are you using the latest code in the 0.8 branch? Thanks, Jun On Tue, May 7, 2013 at 6:48 AM, Yu, Libo wrote: > I got this error when I ran > ./sbt release-zip > > : command not found > [info] Loading project definition from /opt/kafka/project > [info] Set current project to Kafka (in build fi

Re: a few questions about kafka usage

Do you think you can improve the wiki at https://cwiki.apache.org/confluence/display/KAFKA/Performance+testing ? Thanks, Jun On Tue, May 7, 2013 at 6:23 AM, Yu, Libo wrote: > I am using 0.8. I figured out the issue. I used only one thread in each > producer. > That is why the performance degr

Re: slow log recovery

If a broker is down, the cluster will be running in under replicated mode, ie, data will be written to fewer replicas. When the broker comes back, it will catch up data from the current leader. Thanks, Jun On Mon, May 6, 2013 at 10:23 PM, Jason Rosenberg wrote: > Will producers also be able t

Kafka C++ Sync Producer on Windows

Hi All, I am using Kafka 0.7.2. I was able to run the async example of the C++ producer on Windows using the code listed at https://github.com/kafka-dev/kafka/blob/master/clients/cpp/src/producer.cpp But as I understand, the C++ async producer does not have the fail-over mechanisms of bufferin

Re: exception report

Hi, This is possible if the metadata request has several topics. We are working on making it more efficient. For now, you can increase the value of request.timeout.ms. Thanks, Neha On May 7, 2013 6:29 AM, "Yu, Libo" wrote: > Hi, > > This exception was thrown after the server was running for abo

Re: develop customized producer and consumer in java

I got this error when I ran ./sbt release-zip : command not found [info] Loading project definition from /opt/kafka/project [info] Set current project to Kafka (in build file:/opt/kafka/) [error] Not a valid command: release-zip [error] Expected '/' [error] Expected ':' [error] Not a valid key: re

exception report

Hi, This exception was thrown after the server was running for about a day. [2013-05-05 21:53:15,377] WARN Fetching topic metadata with correlation id 0 for topics [Set(atest1)] from broker [id:0,host: ,port:9092] failed (kafka.client.ClientUtils$) java.net.SocketTimeoutException

Re: a few questions about kafka usage

I am using 0.8. I figured out the issue. I used only one thread in each producer. That is why the performance degrades greatly when the number of topics and number of partitions increase. In fact, this number of threads is a very important parameter but not mentioned in the documents. Regards,