seem the pig task launch hang because the yarn.resourcemanager.hostname is not generated into yarn.resourcemanager.scheduler.address property of job.xml generated file on the container slave node.
After it try to connect to adress 0.0.0.0:8030 scheduler instead of stargate:8030 define in yarn-site.xml and cant continu run because by lot of retrying on scheduler connection.
bug 2.6.0 ?