[kepler-users] Error in running Kepler 2.3 on a cluster

Derik Barseghian barseghian at nceas.ucsb.edu
Fri Sep 7 17:30:21 PDT 2012


Glad you got it working, and thanks for reporting what fixed it for you.
I'm not sure why changing our "Min Memory" setting (which controls Xms--initial heap size) fixes it for you, maybe your 64-bit JVM needs more than 5m (although I've been fine with 5m on Mac with 64bit java 1.6).
Anyways, it strikes me our default Xms of 5m is way too small, since Kepler 2.3 at idle after startup seems to use about 300m. Supposably allocating more from the start can benefit performance since the jvm doesn't have to do it. I did a crude stopwatch comparison and didn't notice any startup time decrease using 300m. Regardless I've increased out default to 256m to avoid problems like you saw. This change will go out with Kepler 2.4.

Derik


On Sep 6, 2012, at 11:41 PM, Himanshu ............. <himanshu.king at gmail.com> wrote:

> Hi Derik,
> 
> The problem is solved. Kepler is running on Cluster now (still not running on my Virtual Machine, guess something is wrong with VM itself). I edited the memory.xml file but instead decreasing the max memory, my problem was solved by increasing the min memory (I changed it from 5m to 20m). But still wondering what is wrong with the default memory settings.
> 
> Thanks for your reply
> 
> Himanshu
> 
> On Fri, Sep 7, 2012 at 12:25 AM, Derik Barseghian <barseghian at nceas.ucsb.edu> wrote:
> Hi Humanshu,
> I think I was wrong, kepler 2.3.0 uses 512m by default, not 1000m. 2.4.0 will use 1000m.
> 
> How are you determining the change is not taking affect? I just tried and it worked for me.
> 
> This is what I did for my release version of kepler 2.3.0 on ubuntu:
> vi /home/derik/KeplerData/kepler.modules/build-area/settings/memory.xml
> #change max to something smaller:
> <entry key="max">256m</entry>
> #save, and start kepler
> ./kepler.sh
> 
> # During startup I see the change in a stdout msg:
> Kepler.run going to run.setMain(org.kepler.Kepler)
> JVM Memory = 5m 256m
> 
> If you're using the 2.3 branch instead of 2.3.0 release, you may need to also change this copy of memory.xml:
> vi kepler.modules23/build-area/settings/memory.xml
> 
> Derik
> 
> On Sep 6, 2012, at 9:11 PM, "Himanshu ............." <himanshu.king at gmail.com> wrote:
> 
>> Hi Derik,
>> 
>> Thanks for your reply. Its a virtual machine on which I am trying to run Kepler. When I had first encountered this error,I concluded the same thing so I increased the RAM size but still I got the same error. Since Kepler is not getting started at all, so I can't change the JVM memory settings. And as per your second suggestion I tried editing the memory.xml file but changes are not getting affected.
>> 
>> Thanks
>> Himanshu
>> 
>> On Thu, Sep 6, 2012 at 2:54 PM, Derik Barseghian <barseghian at nceas.ucsb.edu> wrote:
>> Hi,
>> 
>> Kepler 2.3 uses 1gb of memory by default. Maybe the machine on which you're getting the error does not have this much RAM available?
>> You can adjust Kepler's memory use from within Kepler using the Tools=>JVM Memory Settings menu, or by editing the KeplerData/kepler.modules/build-area/settings/memory.xml file.
>> Decreasing the max memory setting to 500m may be fine if you're not running complicated workflows.
>> 
>> Derik
>> 
>> On Sep 6, 2012, at 4:50 AM, "Himanshu ............." <himanshu.king at gmail.com> wrote:
>> 
>>> Dear All,
>>> 
>>> I am trying to run Kepler 2.3 on my cluster. Operating system is Red Hat and Java version is 1.6 (64 bit). I am getting following error.
>>> 
>>> [null] Error occurred during initialization of VM
>>> [null] Too small initial heap for new size specified
>>> 
>>> It looks like a JVM error to me but i am not sure how to change the minimum heap size. If it helps, I have installed Kepler 2.3 on my local machine which is running on fedora and 32 bit Java and it is working fine. I am wondering why the default memory size is not sufficient to start Kepler 2.3 on a cluster or problem is something else. Please give me suggestions to solve this problem.
>>> 
>>> Thanks
>>> _______________________________________________
>>> Kepler-users mailing list
>>> Kepler-users at kepler-project.org
>>> http://lists.nceas.ucsb.edu/kepler/mailman/listinfo/kepler-users
>> 
>> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nceas.ucsb.edu/kepler/pipermail/kepler-users/attachments/20120907/7057f3ce/attachment.html>


More information about the Kepler-users mailing list