spark-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Akhil Das <>
Subject Re: Spark Streaming scheduling control
Date Mon, 20 Oct 2014 08:22:48 GMT
What operation are you performing? And what is your cluster configuration?
If you are doing some operation like groupBy, reduceBy, join etc then you
could try providing the level of parallelism. if you give 16, then mostly
each of your worker will get 8 tasks to execute.

Best Regards

On Mon, Oct 20, 2014 at 3:49 AM, davidkl <> wrote:

> Hello,
> I have a cluster 1 master and 2 slaves running on 1.1.0. I am having
> problems to get both slaves working at the same time. When I launch the
> driver on the master, one of the slaves is assigned the receiver task, and
> initially both slaves start processing tasks. After a few tens of batches,
> the slave running the receiver starts processing all tasks, and the other
> won't execute any task more. If I cancel the execution and start over, the
> roles may switch if the other slave gets to be assigned the receiver, but
> the behaviour is the same, and the other slave will stop processing tasks
> after a short while. So both slaves are working, essentially, but never at
> the same time in a consistent way. No errors on logs, etc.
> I have tried increasing partitions (up to 100, while slaves have 4 cores
> each) but no success :-/
> I understand that Spark may decide not to distribute tasks to all workers
> due to data locality, etc. but in this case I think there is something
> else,
> since one slave cannot keep up with the processing rate and the total delay
> keeps growing: I have set up the batch interval to 1s, but each batch is
> processed in 1.6s so after some time the delay (and the enqueued data) is
> just too much.
> Does Spark take into consideration this time restriction on the scheduling?
> I mean total processing time <= batch duration. Any configuration affecting
> that?
> Am I missing something important? Any hints or things to tests?
> Thanks in advance! ;-)
> --
> View this message in context:
> Sent from the Apache Spark User List mailing list archive at
> ---------------------------------------------------------------------
> To unsubscribe, e-mail:
> For additional commands, e-mail:

View raw message