Hi Jordi, 1. Are you running the job in one machine yarn? or in a cluster?
2. what kind of the java process do you see after killing the yarn application? Because usually, when we run kill-yarn-job applicationId, we do kill all the processes (this is actually done by the Yarn). 3. Which version of Samza and Yarn are you using ? This matters sometimes. Thanks, Fang, Yan yanfang...@gmail.com On Tue, Sep 22, 2015 at 3:42 PM, Jordi Blasi Uribarri <jbl...@nextel.es> wrote: > Hi, > > I am currently developing solution using samza and in the development > process I need to constantly change the code and test in the system. What I > am seeing is that most of the times I kill a job using the kill-yarn-job > script the job gets killed according to the web interface but I see the > java process running. I also have seen that the job was actually been > executed, as I got messages in the far end of the application. I have been > manually killing these processes (kill -9 ) but I have some questions: > > > - Is there a reason for the processes not to be killed. It was > not a matter of time as I could find them hours later. > > - I don’t know if there should be any other action performed to > completely clean the information or killing the process the hard way is > enough. > > - I am finding some memory consumption problems that I don’t know > if they are related with this. Maybe I will describe them in another > message. > > Thnaks, > > Jordi > ________________________________ > Jordi Blasi Uribarri > Área I+D+i > > jbl...@nextel.es > Oficina Bilbao > > [http://www.nextel.es/wp-content/uploads/Firma_Nextel_2015.png] >
