Just now I noticed that BOINC wasn't doing any background processing. Then I also noticed that 11 hours ago, it came up with "unable to stop applications - aborting CPU benchmark". As a result I've lost 11 hours' worth of processing. I'm using version 5.4.11.

Can anybody shed any light on this please? It's not the first time it has happened. Is it a known issue?

---

I have no idea.

What about upgrading to the latest stable? 5.4.11 is pretty old now.

---

Kathryn :o)

Yes, it's a known issue.

As you noticed, it happens when benchmarks are aborted because an application that has to be removed from memory failed to respond to the quit request within the 10 second interval that the BOINC core client allows. The problem is that the delayed quit response is never acted on and the core client keeps the application marked as running. If memory serves me right things do sort themselves out if the core client decides to schedule another result. It affects applications that have sequences of calculations that are indivisible (e.g. CPDN post-processing) most frequently.

It was first raised on the development mailing list more than 2 years ago by Chris Sutton and myself. I can remember raising it on the list at least twice since then with suggested fixes.

Edit: just checked and the 10 second limit is still in place for 5.10.*.

---

"The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer

Yes, it's a known issue.

As you noticed, it happens when benchmarks are aborted because an application that has to be removed from memory failed to respond to the quit request within the 10 second interval that the BOINC core client allows. The problem is that the delayed quit response is never acted on and the core client keeps the application marked as running. If memory serves me right things do sort themselves out if the core client decides to schedule another result. It affects applications that have sequences of calculations that are indivisible (e.g. CPDN post-processing) most frequently.

It was first raised on the development mailing list more than 2 years ago by Chris Sutton and myself. I can remember raising it on the list at least twice since then with suggested fixes.

Edit: just checked and the 10 second limit is still in place for 5.10.*.

Thanks Thyme. I upgraded as also suggested, but from what you say this seems to be a known and as yet unresolved issue. And yes, it was CPDN that was involved. I resolved it by restarting BOINC manager.

---

Well, one thing did change between 5.4 and 5.10 and that is that when benchmarks are run, the application and result stays in memory. In 5.4 it would be unloaded from memory prior to running benchmarks, yet because this caused problems with applications that didn't checkpoint and ran huge work units which would restart from zero percent after the benchmark run, it was changed that the application stays in memory during the benchmarks.

So in other words, although the CPU cycles will go to the benchmarking process, the application won't stop in 5.10

---