-
Getting containers that die to automatically restart
-
Get the Application Master to restart if it dies
-
Management of the clients. Currently I have to kill clients through the YARN Cli.
`yarn application -kill [app#] `
-
Adding in unit tests and sample/test applications
-
Have TOYA clean things up if the AM dies or is exited.
-
Migrate paramaters to configeration file
-
Add Jetty server to handle commands sent to the AM
-
Develop Zookeeper watcher to keep load balancer in sync with running tomcat servers
-
Add option to toggle remote debugging
Password-less ssh to sandbox is setup: cat ~/.ssh/id_rsa.pub | ssh [email protected] -p 2222 'cat >> .ssh/authorized_keys'
hadoop jar [toya-client jar] org.toya.core.yarn.Client
usage: Client
-appname <arg> Optional: Application Name. Default value - toya
-container_memory <arg> Recommended: Amount of memory in MB to be requested to run
the shell command - Defaults to 10, Recommended is 512.
-debug Optional: Dump out debug information
-help Optional: Print usage
-jar <arg> Required: Jar file containing the application master - toya-client jar
-lib <arg> Required: Runnable Jar with toya inside - toya-server jar
-log_properties <arg> Optional: log4j.properties file
-master_memory <arg> Recommended: Amount of memory in MB to be requested to run
the application master - Defaults to 10, Recommended is 128
-toya_priority <arg> Optional: Priority for the TOYA containers - Defaults to 0
-num_containers <arg> Recommended: No. of containers on which the shell command
needs to be executed, Defaults to 1
-priority <arg> Optional: Application Priority - Default 0
-queue <arg> Optional: RM Queue in which this application is to be
submitted - Defaults to 'default'
-ZK <arg> Required: Comma seperated list of ZK hosts ie -
host1:port,host2:port
-webapp_root <arg> Required: Webapp root on HDFS
-start_port <arg> Required: Start of tomcat port range on each node
See start.sh script.