> On 21 May 2015, at 02:25, Rajesh Kartha <[email protected]> wrote: > > Thanks a lot Steve, > > So the suggestion is to use the AMRMClient.updateBlacklist() method to > include *nodes where the component is running + unreliable nodes* before > the container is requested instead of discarding allocated containers. >
that's right. 1. We'd have to include in the blacklist those nodes on which instances have already been explicitly requested but not yet granted (outstanding request tracker knows them) 2. and ask for each instance one by one. Otherwise if you ask for >1 container they can be assigned to the same host > I will take a closer look and update SLIDER-82 with my approach. > > Thanks, > Rajesh > that'd be great!
