To run Tigon in standalone mode:
$ run_standalone.sh <path-to-flow-jar> <flow-class-name> <run-time-args>
$ run_standalone.sh DataParser.jar com.example.TopFlow --key1=value1 --key2=value2
The Flow program will be started and run in the same Java process. Only one Flow can be run at a time, and the number of Flowlet instances cannot be changed once started. Runtime arguments are dependent on the application being run.
To stop the application, its Flow and Flowlets, quit the shell script using control-C.
To run Tigon in distributed mode:
$ run_distributed.sh <zookeeper-quorum> <hdfs-namespace>
$ run_distributed.sh example.com:2181/quorum tigon
The distributed mode will start the Flow. Flowlet instances will be spun up as YARN containers in the Hadoop cluster.
The Flow will appear as a YARN Application and will be visible through the YARN UI, located at http://<master-node>:8088. For the example above, that would be:
The distributed mode has a command-line interface that can be used to control the Flow and its lifecycle. A complete list of available commands is documented in the Tigon Tools; to list the running Flows, you can use:
tigon > list
You can list all the running Flows, start and stop a Flow, check the status of a running Flow, and delete all queues associated with a particlar Flow.
Note that, in distributed mode, Flows continue to run even after you quit the command-line interface unless you have explicitly stopped them. You can stop and delete a particluar Flow with the command:
tigon > delete <flow-name>
You can then exit the command-line interface either with exit or Control-C.