A typical leak is creating a context with ContextBuilder, but never closing it. Could that be the case here?
With latest jclouds (and default http driver), the only use of threads are aggregate operations like clearing a container. On Nov 18, 2014 8:34 AM, "Daniel Hsueh" <[email protected]> wrote: > Thanks Adrian! > > In the meantime, I have some more observations. The old-gen GC > correlation occurs with incgc (4G Xmx and Xms), cms (4G), and G1 (2G and > 1G); G1 with 4G memory sees failures not correlated with any old-gen run > (no old-gen runs occurred at all). > > The glut of threads are all showing up in jconsole as > http-bio-8080-exec-nn where nn is a number and 8080 is the endpoint port > number. The numbers increase every approx 4.5 minutes and most go away > after about 10min idle time (10 threads with nonconsecutive nn values stay > behind). > > I'll see if I can get an auth-v2 endpoint set up for the same blobstore > (to use the openstack-swift provider). Are there any ideas I should try in > the meantime? > > Thank you. > > On Mon, Nov 17, 2014 at 3:26 PM, Adrian Cole <[email protected]> > wrote: > >> Ok. I will add it :) >> > > > > -- > Daniel Hsueh -- mailto:[email protected] <[email protected]> > tel:+1-905-287-2167 >
