And also gradually reduce the number of tasks run via the "use at most % processors" preference?

Is this a work around just to not have it fixed? I just stop getting new task from TLP and ATLAS. I rather have all cores running. Thank you but I can stop it for the time being this way.

This is to check if a project application maybe responsible for the hangs or the Client overtasking the memory by starting tasks that should not be started. It is not a workaround but would help us to locate the problem.

Just as a sidenote we once had the case of an alleged memory leak in the vboxwrapper (the same app ATLAS@home uses to run VM tasks) where the memory usage increased over time (linear to the amount of VM tasks running) and freezed the system when no RAM was available. The investigation took some time becaue we could reproduce this only on this one system of the volunteer. In the end it turned out to be a driver issue on the host. The volunteer was using an SSD but installed a driver that was designed for HDD (spinning disks) which had the unwanted sideeffect that the Windows kernel didn't free memory of the VirtualBox application which lead to the increase in memory usage. Something that seemed completely unrelated was causing the issue. I want to make sure we also cover all areas in this case.

Can't the application be told via app_config <max_concurrent> to use less core resources and achieve same, so rest can be used for small project tasks?

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

Can't the application be told via app_config <max_concurrent> to use less core resources and achieve same, so rest can be used for small project tasks?

Yes, this way one can reduce the tasks of a specific project while still using all cores. This involves fiddling with config files which I don't think is necessary at this point of the discussion.

Maybe we should also try another angle and monitor the memory consumption of the system via resource monitor. This helped to identify the problem in the case I described earlier. But this can only be done while in front of the computer.

Could someone running ATLAS@Home tell me

1. rsc_memory_bound for an ATLAS task
2. memory_size_mb from ATLAS's vbox_job.xml file (the actual file name may include version number and such)