Hi Safder, I have never seen that behavior. I just tested your scenario, and it does not happen for me.
Twill will only attempt to start another instance of a runnable if that runnable has failed. Does that first application (the runnable) exit with an error? Or perhaps call System.exit()? You would see something like this in the logs: 2014-03-28T20:08:06,162Z WARN o.a.t.i.a.RunningContainers [my.host.name] [ServiceDelegate] RunningContainers:handleCompleted(RunningContainers.java:314) - Container container_1395888520219_0009_01_000002 exited abnormally with state COMPLETE, exit code 1. -Andreas. On Fri, Mar 28, 2014 at 12:35 PM, safder <[email protected]> wrote: > Hi Guys, > > Need some help again. I am trying to understand application life cycle. I > have 2 different applications having very different run times i.e. once > application can finish well before the other. My simple tests seem to > indicate, that if one application finishes before another, twill tries to > recreate another instance of the first application. At least thats what I > see using the TwillRunnerService in the YarnTestUtils. (i.e. the testing > cluster) > > Can someone shed a light on this for me? > > Thanks > Safder..
