Hey Suresh! The old way of setting the memory limits via mapreduce properties should work after OOZIE-2896 <https://issues.apache.org/jira/browse/OOZIE-2896> See that jira for some details. However the new - and preferred - way of doing so is to add a launcher configuration to the action like:
<launcher> <memory.mb>4096</memory.mb> </launcher> This should go between the name-node and the job-xml tags or can go into the workflow's <global> section to apply to every action in there. See the Common schema <https://oozie.apache.org/docs/5.0.0/WorkflowFunctionalSpec.html> here for some details. The third option is to increase the default by setting the oozie.launcher.default.memory.mb in the oozie-site.xml. Hope it helps, gp On Wed, Aug 29, 2018 at 10:02 PM Suresh V <verdi...@gmail.com> wrote: > We recently launched an EMR cluster with latest version of Oozie that is > Oozie 5.0.0. > > We understand the Oozie launcher is no more a mpareduce job in Yarn. > We see that it shows up as 'Oozie launcher' in the Yarn UI. > > Our workflow has a Spark and a Sqoop job, and the launcher is failing first > attempt with running out of memory error, causing it to try multiple > attempts. > Please advise where we can set the memory limits for Oozie launcher, in > order to work around this error? > > AM Container for appattempt_1534385557019_0066_000001 exited with exitCode: > -104 > Failing this attempt.Diagnostics: Container > [pid=19109,containerID=container_1534385557019_0066_01_000001] is running > beyond physical memory limits. Current usage: 2.2 GB of 2 GB physical > memory used; 9.9 GB of 10 GB virtual memory used. Killing container. > > > Thank you > Suresh. > -- *Peter Cseh *| Software Engineer cloudera.com <https://www.cloudera.com> [image: Cloudera] <https://www.cloudera.com/> [image: Cloudera on Twitter] <https://twitter.com/cloudera> [image: Cloudera on Facebook] <https://www.facebook.com/cloudera> [image: Cloudera on LinkedIn] <https://www.linkedin.com/company/cloudera> ------------------------------