04-06-2016 09:47 AM
on CDH 5.X seems it's not possible to create an autoplacement rule based on secondary group if it's using also the subpool.
In the cloudera documentation it's possible to create placement rule based on root.<SecondaryGroupName>.
When you create a sub queue root.poolparent.pool1 the pool name searched by the rule is poolparent.pool1, but
in Apache Yarn Documentation ( https://hadoop.apache.org/docs/r2.7.1/hadoop-yarn/hadoop-yarn-site/FairScheduler.html#Queue_Access_C... ) shows that a group name with dot "." is translated with "_dot_" . So it's not ok to create a group poolparent.pool1 because it will be translate in "poolparent_dot_pool1"