While not an answer to your question I experienced this problem when I was building a native YARN app. Zookeeper solved the issue of 'host discovery' for me very easily. By having each 'daemon' in a container join a zookeeper node while its alive gives clients the ability to connect to the ZK cluster and then discover were the 'daemon' hosts are running.
Again while not slider this is done in MOYA. http://hortonworks.com/blog/how-to-deploy-memcached-on-yarn/ On Mon, Jul 7, 2014 at 2:42 PM, Rui Zhang <rzh...@vertica.com> wrote: > Hi, > > I am now working on make a tarball for my own application. What I mean is > start the application in the script like the HBase example you gave to us. > In the python script, we need some commands for each role. If the command > of my application can start more than one processes, does that mean all the > processes are running in just one container? > > My guess is that the python script is run on the machine of that specific > container. So the processes initiated by the script should run in this > container. Am I correct? > > Rui > > > On 07/07/2014 01:37 PM, Steve Loughran wrote: > >> On 7 July 2014 17:34, Rui Zhang <rzh...@vertica.com> wrote: >> >> Hi, >>> >>> How can I know the ip address or hostname of containers after they are >>> allocated? Because if I want to start an distributed application, I need >>> to >>> set the host names. >>> >>> YARN chooses where to run them -you can get the list of the containers >> via >> the web UI/REST API once they are up. >> >> >> And if I start all the instances in one command, is it ensured that all >>> the instances will run in containers allocated by Yarn? >>> >>> >>> what do you mean by "start all the instances?" If it is just >> starting/thawing a slider cluster, then every component is always >> requested >> in its own YARN container. >> >> One thing we do have to deal with is getting configurations back to client >> machines when the locations of services are not known in advance. We have >> a >> zookeeper based service registry for this, with a web view of it on the >> slider AM -and I'm working to do a YARN-wide registry which will work for >> all YARN-deployed applications >> >> -steve >> >> > -- > Rui Zhang > Software engineer Intern > Vertica, an HP Company > rzh...@vertica.com > > -- Hortonworks Certified Hadoop 2 Java Developer Solutions Engineer, Hortonworks Phone: Office (734)-288-3930 Mobile (734)-301-9812 Email: jniem...@hortonworks.com Apache Hadoop YARN: Moving beyond MapReduce and Batch Processing with Apache Hadoop 2 <http://www.amazon.com/Apache-Hadoop-YARN-Processing-Addison-Wesley/dp/0321934504> -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.