On 16.05.2014, at 03:19, Michael Dürig <[email protected]> wrote:

> Turns out that this is a memory consumption issue of short lived sessions 
> causing the system to degrade due to memory pressure. Java 7 and 8 use larger 
> heap sizes by default so we are not seeing this there. Running the test with 
> -Xmx1024M produces comparable results across all Java versions.

Ah, great, I kind of missed to make that test with the same heap size.

> The issues is caused by RegistrationCallable apparently deferring gc until 
> the callable have been called. In a quick test I removed this part of the 
> code, which considerably lowered memory consumption of the test run.
> 
> Will follow up with an issue.

For the record, this was https://issues.apache.org/jira/browse/OAK-1833

Thanks,
Alex

Reply via email to