The AM is a special container in each YARN application that requests resources for the application and monitors them once they are allocated (see https://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/YARN.html for general information about YARN). If just the AM goes down, YARN will restart it. In Slider's implementation of an AM, the AM will usually continue where it left off. But there are still things that can make the entire application go down. Slider has a configurable parameter for how many container failures it will allow before it stops the application.
Are you using some kind of custom AM? On Mon, Mar 6, 2017 at 2:53 PM, jeff saremi <jeffsar...@hotmail.com> wrote: > Thanks Billie, > > has anyone come up with anything that would monitor the AM itself? I still > dont' know what it is but i'm guessing its some sort of application manager > in Slider. > > ________________________________ > From: Billie Rinaldi <billie.rina...@gmail.com> > Sent: Monday, March 6, 2017 12:27:18 PM > To: dev@slider.incubator.apache.org > Subject: Re: Newbie Question: Can Slider detect application exit and > restart if needed? > > The Slider AM for each app takes care of restarting containers that have > exited, but if the YARN app itself has stopped, the AM will no longer be > running and the app will not be restarted automatically. > > On Mon, Mar 6, 2017 at 11:04 AM, jeff saremi <jeffsar...@hotmail.com> > wrote: > > > We're currently having issues with our Yarn jobs where they exist without > > our knowledge. > > is there anyway that Slider alleviates this concern? > > thanks > > > > Jeff > > >
