spark-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Matei Zaharia <>
Subject Re: Mesos/Spark Deadlock
Date Mon, 25 Aug 2014 04:52:00 GMT
Yeah, Mesos in coarse-grained mode probably wouldn't work here. It's too bad that this happens
in fine-grained mode -- would be really good to fix. I'll see if we can get the workaround
in into Spark 1.1. Incidentally have you tried


On August 23, 2014 at 4:30:27 PM, Gary Malouf ( wrote:

Hi Matei,

We have an analytics team that uses the cluster on a daily basis.  They use two types of
'run modes':

1) For running actual queries, they set the spark.executor.memory to something between 4 and
8GB of RAM/worker.  

2) A shell that takes a minimal amount of memory on workers (128MB) for prototyping out a
larger query.  This allows them to not take up RAM on the cluster when they do not really
need it.

We see the deadlocks when there are a few shells in either case.  From the usage patterns
we have, coarse-grained mode would be a challenge as we have to constantly remind people to
kill their shells as soon as their queries finish.  

Am I correct in viewing Mesos in coarse-grained mode as being similar to Spark Standalone's
cpu allocation behavior?

On Sat, Aug 23, 2014 at 7:16 PM, Matei Zaharia <> wrote:
Hey Gary, just as a workaround, note that you can use Mesos in coarse-grained mode by setting
spark.mesos.coarse=true. Then it will hold onto CPUs for the duration of the job.


On August 23, 2014 at 7:57:30 AM, Gary Malouf ( wrote:

I just wanted to bring up a significant Mesos/Spark issue that makes the
combo difficult to use for teams larger than 4-5 people. It's covered in My understanding is that
Spark's use of executors in fine-grained mode is a very different behavior
than many of the other common frameworks for Mesos.

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message