----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/45220/#review125069 -----------------------------------------------------------
ambari-server/src/main/resources/common-services/YARN/2.1.0.2.0/package/scripts/yarn.py (line 75) <https://reviews.apache.org/r/45220/#comment187811> Shouldn't this just be create direcotry with parents allowed rather than checking explicitly for /tmp? - Alejandro Fernandez On March 23, 2016, 4:40 p.m., Laszlo Puskas wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/45220/ > ----------------------------------------------------------- > > (Updated March 23, 2016, 4:40 p.m.) > > > Review request for Ambari, Andrew Onischuk, Sumit Mohanty, and Sebastian > Toader. > > > Bugs: AMBARI-15531 > https://issues.apache.org/jira/browse/AMBARI-15531 > > > Repository: ambari > > > Description > ------- > > When cluster created via blueprint the /tmp folder may be created implicitly > by various components with permissions that prevent other components to write > to it. In the specific case described by the linked issue, the folder has > been created when starting the historyserver, and later on the folder > couldn't be written by the hiveserver2. (ideally the folder is expected to be > created when the namenode starts) > > The patch fixes the specific case described in the bug, however a more > robust/generic solution is needed to properly sort out the problem. > > > Diffs > ----- > > > ambari-server/src/main/resources/common-services/YARN/2.1.0.2.0/package/scripts/yarn.py > 05e19cf > > Diff: https://reviews.apache.org/r/45220/diff/ > > > Testing > ------- > > Unit tests in progress. > Manual testing under way > > > Thanks, > > Laszlo Puskas > >