> On June 6, 2016, 4:21 p.m., Sumit Mohanty wrote: > > ambari-agent/conf/unix/ambari-agent.ini, line 37 > > <https://reviews.apache.org/r/48207/diff/3/?file=1407014#file1407014line37> > > > > Whats the default behavior - looks like by default it should be > > disabled. > > > > What is the memory consumption for an agent instance that is not > > leaking - assuming 20+ master/slave components on the host.
done - Andrew ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/48207/#review136283 ----------------------------------------------------------- On June 6, 2016, 4:29 p.m., Andrew Onischuk wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/48207/ > ----------------------------------------------------------- > > (Updated June 6, 2016, 4:29 p.m.) > > > Review request for Ambari and Dmitro Lisnichenko. > > > Bugs: AMBARI-17030 > https://issues.apache.org/jira/browse/AMBARI-17030 > > > Repository: ambari > > > Description > ------- > > Add an agent capability to auto-restart when the memory threshold for the > agent process exceeds a limit. This should be a configurable property (e.g. > agent_memory_threshold) where any positive value, say in MB, is a limit at > which agent should get restarted. > > The restart itself should be executed when there are no > EXECUTION/AUTO_EXECUTION commands in pipeline. So when the condition is > reached for agent auto restart, a flag should be set to restart when agent is > not doing any work. > > A hard limit should also be set where the agent will be restarted even if > there are commands in pipeline. > > > Diffs > ----- > > ambari-agent/conf/unix/ambari-agent.ini aacbb8a > ambari-agent/src/main/python/ambari_agent/Controller.py 91bc586 > ambari-common/src/main/python/ambari_commons/os_utils.py 4a636d9 > > Diff: https://reviews.apache.org/r/48207/diff/ > > > Testing > ------- > > mvn clean test > > > Thanks, > > Andrew Onischuk > >