This commercial erp is highly customised downstream project of ofbiz 10.04, OT actually. I profiled the app, identified changes in the framework changes and ensured I worked them back into the code, a bridged port if you will. I looked at the solution ofbiz applied to these very same threading issues I found. But framework/start was taken directly from ofbiz 13, nothing needed to be changed because this part of framework was not customized. Most of the other customization to framework are superficial.
On Sat, Jul 9, 2016 at 6:19 PM, Taher Alkhateeb <[email protected]> wrote: > Hi Justin, > > What do mean by project version 10? Do you mean OFBiz 10? > > Also do you have logs to show the progression towards dead locks? This > issue seems extremely technical with backported changes from two releases > back an old unsupported release. It would almost be impossible to help you > out without some context and background details. > > FYI the class below is an implementation of OFBiz specific classloader that > adds stuff to the classpath. If you made customizations to the classpath > then that might be a reason. > > HTH > > Taher Alkhateeb > On Jul 9, 2016 2:20 PM, "Justin Robinson" < > [email protected]> wrote: > > > I'm profiling a downstream project version 10, I recently > > replaced/refactored/bridged the framework components with obiz 13 and > some > > obiz 14 (just the transaction related in entity, to use dbcp2 etc. > > > > Anyway the monitor deadlocks I saw with initial testing on the version 10 > > based multitenant erp system are all gone and in there place is this one > > monitor. In jmeter tests of 5000 users spread accross 200 tenants I'm > > getting 100% failure and all threads are blocked and time-out as a result > > > > 5 000 http-bio-0.0.0-8080-exec-xxx deadlock on monitor > > org.base.start.Classpath$NativeLibClassLoader > > >
