-
Bug
-
Resolution: Incomplete
-
P3
-
None
-
8u72
-
x86
-
windows_2012
FULL PRODUCT VERSION :
This happens on 8u71 or 8u72
Problem was resolved by downgrading to 8U25
ADDITIONAL OS VERSION INFORMATION :
6.3 9600 Server 2012 R2
EXTRA RELEVANT SYSTEM CONFIGURATION :
This is a G9 with two sockets each of 18 cores - set up for Hyperthreading
A DESCRIPTION OF THE PROBLEM :
We have a Java process launching many executables.
These "engines" happen to be written in C++ with one main working thread each. They are designed to perform CPU and data intensive workloads.
The hava process that launches this process is simply a broker that hands off contol to other process running on the same or different boxes. The problem
happens irrespective of this.
On the latest Java it is only possible ot max out one physical processor.
on 8u25 the workload is spread evenly across both CPUs allowing the box to
be maxed out.
Checked all the bios settings - NUMA etc - no difference.
REGRESSION. Last worked in version 8u60
ADDITIONAL REGRESSION INFORMATION:
Actually 8U25. It may work in intermediate versions - I have not tested
STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Use java to launch CPU intensive executables on a multi core box with more than one processor
EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
. See if all child processes run on one side or not
ACTUAL -
Can ony max out one CPU on the box.
ERROR MESSAGES/STACK TRACES THAT OCCUR :
No crash
REPRODUCIBILITY :
This bug can be reproduced always.
CUSTOMER SUBMITTED WORKAROUND :
Downgrate to 25 which is now quite an old version.
This happens on 8u71 or 8u72
Problem was resolved by downgrading to 8U25
ADDITIONAL OS VERSION INFORMATION :
6.3 9600 Server 2012 R2
EXTRA RELEVANT SYSTEM CONFIGURATION :
This is a G9 with two sockets each of 18 cores - set up for Hyperthreading
A DESCRIPTION OF THE PROBLEM :
We have a Java process launching many executables.
These "engines" happen to be written in C++ with one main working thread each. They are designed to perform CPU and data intensive workloads.
The hava process that launches this process is simply a broker that hands off contol to other process running on the same or different boxes. The problem
happens irrespective of this.
On the latest Java it is only possible ot max out one physical processor.
on 8u25 the workload is spread evenly across both CPUs allowing the box to
be maxed out.
Checked all the bios settings - NUMA etc - no difference.
REGRESSION. Last worked in version 8u60
ADDITIONAL REGRESSION INFORMATION:
Actually 8U25. It may work in intermediate versions - I have not tested
STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Use java to launch CPU intensive executables on a multi core box with more than one processor
EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
. See if all child processes run on one side or not
ACTUAL -
Can ony max out one CPU on the box.
ERROR MESSAGES/STACK TRACES THAT OCCUR :
No crash
REPRODUCIBILITY :
This bug can be reproduced always.
CUSTOMER SUBMITTED WORKAROUND :
Downgrate to 25 which is now quite an old version.