Hi, In Zeppelin side,
Depending on what scheduler is returned from individual interpreter implementation "Interpreter.getScheduler()", Zeppelin runs job sequentially or concurrently. For example, if interpreter implementation returns FIFO scheduler, paragraphs using the interpreter will run one by one, sequentially. If interpreter implementation return Parallel scheduler, paragraphs using the interpreter will run concurrently. In Spark Interpreter, Basically, SparkInterpreter implementation uses FIFO scheduler. Because of code passed to scala / python compiler need to have an order. But SparkContext it self is thread safe and it's got fair scheduler inside. So Zeppelin's SparkSQLInterpreter leverage fair scheduler of Spark to execute query concurrently. Set zeppelin.spark.concurrentSQL true will make SparkSQLInterpreter use Parallel scheduler, and it allow multiple paragraphs running concurrently, and Spark processes the job with it's fair scheduler. So, Zeppelin is capable of running job concurrently. It depends on which interpreter you use and which scheduler is returned from that particular interpreter. Hope this helps. Best, moon On Thu, Nov 19, 2015 at 11:22 AM Fengdong Yu <[email protected]> wrote: > Hi Dev Team, > > I want to confirm with you guys: > > 1) Does that Zeppelin only run one Job concurrently? I mean if I submit > a job on my notebook, so other’s submit will be pending until my job > finished? > 2) How can I submit more than one job at the same time on my notebook? > > > > > >
