perfSONAR User Q&A and Other Discussion

[Andrew Lake <>]

On April 12, 2017 at 4:14:18 AM, Brian Candler () wrote:

On 10/04/2017 19:46, Andrew Lake wrote: 
> 
> There are a a whole bunch of new features in pScheduler vs 
> BWCTL+regular testing. We actually keep a schedule in a database now, 
> so unlike BWCTL, you can know when things ran, what’s running and when 
> they are going to run. We also keep lots more diagnostics information 
> so it’s easier to debug things when a test is missed, etc. In addition 
> to that, it has a plug-in architecture for writing new tests, tools 
> and archivers. 
> 
Thanks - that's very useful to know. 
> 
> First of all, both 2GB and running on a VM would be against what we 
> recommend for 3.5 (and probably a version or two earlier than that as 
> well). That being said I have 2 ESnet productions hosts upgraded to 
> 4.0 running 2GB memory that are fine with a local MA, but they also 
> run a handful of tests and I have optimized cassandra not to eat as 
> much memory at the cost of running bit slower (fwiw I did this 
> optimization when 3.5 was released, because it worked poorly otherwise). 
> 
Is there a doc which summaries the cassandra tweaks suggested? I've now 
increased these VMs to 3GB but would still like it to make better use of 
that memory. 

Buried in our FAQs is a link to this page: http://docs.datastax.com/en/cassandra/2.0/cassandra/operations/ops_tune_jvm_c.html

Basically you can tweak the MAX_HEAP_SIZE and HEAP_NEWSIZE variables in /etc/cassandra/conf/cassandra-env.sh. It was largely trial and error when I did it, but settled on MAX_HEAP_SIZE=512MB and HEAP_NEWSIZE=256MB. Seemed to work ok in my case, but sure it depends on machine load and otehr stuff, so likely you might have to tweak those some. 

Regards, 

Brian.