Hi Rafal! Have you been able to launch the job successfully first without configuring node-labels? Do you really need node-labels? How much total memory do you have on the cluster? Node labels are usually for specifying special capabilities of the nodes (e.g. some nodes could have GPUs and your application could request to be run on only the nodes which have GPUs)
HTH Ravi On Tue, Nov 8, 2016 at 6:19 AM, Rafał Radecki <radecki.ra...@gmail.com> wrote: > Hi All. > > I configured yarn to use capacity scheduler, I have for physical nodes. On > first of them I run resourcemanager and nodemanager on all of them. > > My capacity-scheduler.xml and yarn-site.yml are attached. > When I submit a job to the "long" queue I get in resourcemanager's logfile > the content of attached rm.log and in RM qui the job is in state > "ACCEPTED: waiting for AM container to be allocated, launched and > register with RM" > and has finalstatus > "UNDEFINED" > > At the same time I see that in RM gui in scheduler section I have four > partitions (node1-4d) in which in every one there are two queues "long" and > "short" available. To summarize when I run a task, samza task in my case, I > only specify (https://samza.apache.org/learn/documentation/0.10/jobs/ > yarn-jobs.html) > yarn.queue=long > or > yarn.queue=short > > Have I missed something? > > BR, > Rafal. > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: user-unsubscr...@hadoop.apache.org > For additional commands, e-mail: user-h...@hadoop.apache.org >
