Re: Starting Flink cluster and running a job

2019-02-22 Thread Boris Lublinsky
Gordon, I double checked it several times. Here is the list Columns are [info] - Jar-Size including dependencies [info] - Jar-Size [info] - Number of transitive dependencies [info] - Number of direct dependencies [info] - ModuleID [info] Done updating. [info]TotSizeJarSize #TDe #Dep M

Re: Starting Flink cluster and running a job

2019-02-21 Thread Boris Lublinsky
The relevant dependencies are val flinkScala= "org.apache.flink" %% "flink-scala"% flinkVersion % "provided" val flinkStreamingScala = "org.apache.flink" %% "flink-streaming-scala" % flinkVersion % "provided" val fl

Re: Starting Flink cluster and running a job

2019-02-20 Thread Konstantin Knauf
Hi Boris, can you the relevant parts (dependencies) of your pom.xml? Did you also try without fixing the Kafka version, i.e. running with the Kafka client version provided by the Kafka connector of Flink? Gordon (cc) dealt with FLINK-8741. @Gordon: have you seen this issue with 1.6/1.7 before? C

Re: Starting Flink cluster and running a job

2019-02-20 Thread Boris Lublinsky
I found some more details on this The same error for the same application was reported about a year ago http://mail-archives.apache.org/mod_mbox/flink-user/201802.mbox/%3CCAE7GCT4pF74LwyY=tivzhquq50tkjjawfhaw+5phcsx+vos...@mail.gmail.com%3E

Re: Starting Flink cluster and running a job

2019-02-19 Thread Boris Lublinsky
Thanks Ken, That was my first instinct as well, but.. To run on the cluster I am building an uber jar for which I am fixing Kafka clients jar version I am also fixing version of Kafka So I do not know where another version can get from Boris Lublinsky FDP Architect boris.lublin...@lightbend.com

Re: Starting Flink cluster and running a job

2019-02-19 Thread Ken Krugler
Hi Boris, I haven’t seen this exact error, but I have seen similar errors caused by multiple versions of jars on the classpath. When I’ve run into this particular "XXX is not an instance of YYY" problem, it often seems to be caused by a jar that I should have marked as provided in my pom. Tho

Re: Starting Flink cluster and running a job

2019-02-19 Thread Boris Lublinsky
Konstantin, After experimenting with this for a while, I got to the root cause of the problem I am running a version of a Taxi ride travel prediction as my sample. It works fine in Intellij, But when I am trying to put it in the docker (standard Debian 1.7 image) It fails with a following error

Re: Starting Flink cluster and running a job

2019-02-19 Thread Konstantin Knauf
Hi Boris, without looking at the entrypoint in much detail, generally there should not be a race condition there: * if the taskmanagers can not connect to the resourcemanager they will retry (per default the timeout is 5 mins) * if the JobManager does not get enough resources from the ResourceMan

Starting Flink cluster and running a job

2019-02-17 Thread Boris Lublinsky
Following https://github.com/apache/flink/tree/release-1.7/flink-container/docker I have created an entry point, which looks like follows: #!/bin/sh #