Mixing 1.1.0/1.1.1 should not matter because we didn't touch any YARN features(though I haven't tried it). You didn't mention what kind of setup this is in terms of a cluster size, etc.
XD on YARN did a huge change from 1.0 to 1.1 where whole YARN functionality was pretty re-written from scratch. With 1.0 you had to run separate YARN apps to start admin and containers, with 1.1 this was changed so that both admin and container could be controlled by a same application master.
Having said that, to give a little bit of background, with 1.1, on default(defined in servers.yml) YARN starts 3 containers and 1 admin. If you by any change are working with single node hadoop, you may hit resource limits. It's worth to read section "Configuring YARN memory reservations" from 1.1 reference docs. I've surely burned few time by trying to figure out why something doesn't start on YARN.
Problem might be 1) cluster doesn't have enough resources if those 3 default containers took it all, 2) admin is requestion too much memory and YARN is refusing to fulfil resource allocation request, etc.
What I can propose is to first drop container count to 1 and see if that helps. Secondly try to tweak or turn off memory settings mentioned in that ref doc section.
I've also seen cases with some OS's that ratio of physical mem vs. virtual mem of a process is just calculated wrong by OS(think some versions of RHEL do this) and YARN just kills your process.
It's also worth to try to see from logs if admin was actually started but killed by YARN(due to mem issues), if it failed to start(something went wrong with zookeeper, etc), of if it never got resources from YARN.
Let us know how it goes!