Nope, not a memory problem. I have plenty available on my Windows X64 box.

hi,

The issue has resurfaced after I updated the lates Boinc having VM 5.0.10. This time in the Virtualbox, i am seeing all tasks as Aborted and in Boinc the status as Postponed.

Regards,
Kiran George B.

hi, same too here.

I update boinc version 7.6.22 + virtualbox version 5.0.10. (Windows 7 )

Now camb boinc docker NOT work. run ten minutes and then hypervision failed !!

Older version boinc+vw run ok. vlhc and atlas@home NOT work too. Must return older boinc and vm.

OK, thanks everyone. So its looking like its only happening with Virtualbox 5. Can anyone else check that downgrading to Virtualbox 4 fixes the problem?

Also, it seems to be a Windows-only problem, or can anyone on Linux/Mac disprove this theory?

Also, it seems to be a Windows-only problem, or can anyone on Linux/Mac disprove this theory?

I'm running multiple Linux 64 bit machines with VBox 5.0.2 with no problems. BOINC 7.4.25.

I also have 1 Win7 64 bit machine with VBox 5.0.10 that is running without issue. BOINC 7.6.9.

Hi,

It could be a Windows 10, Virtualbox 5.0.10 issue. That is my configuration and I am facing the issue.

Regards,
Kiran George B.

I am having the problem on Windows 7, BOINC 7.6.22, VirtualBox 5.0.10. The issue started when I upgraded to this BOINC/VirtualBox package. All was fine on the previous version, BOINC 7.6.9 and it's associated VirtualBox (unknown version but I thought it was an earlier v5).

Update: might have found it...apparently v4.3.12. I don't think you can mix different BOINC/VB combinations.

Failed task data :



<core_client_version>7.6.22</core_client_version>
<![CDATA[
<message>
aborted by user
</message>
<stderr_txt>
2016-01-11 17:52:26 (5912): vboxwrapper (7.7.2617): starting
2016-01-11 17:52:26 (5912): Feature: Checkpoint interval offset (542 seconds)
2016-01-11 17:52:26 (5912): Detected: VirtualBox COM Interface (Version: 5.0.10)
2016-01-11 17:52:26 (5912): Detected: Minimum checkpoint interval (600.000000 seconds)
2016-01-11 17:52:26 (5912): Create VM. (boinc_00d1dfab3614e292, slot#4)
2016-01-11 17:52:26 (5912): Updating drive controller type and model for desired configuration.
2016-01-11 17:52:26 (5912): Setting Memory Size for VM. (2048MB)
2016-01-11 17:52:26 (5912): Setting CPU Count for VM. (3)
2016-01-11 17:52:26 (5912): Setting Chipset Options for VM.
2016-01-11 17:52:26 (5912): Setting Boot Options for VM.
2016-01-11 17:52:26 (5912): Enabling VM Network Access.
2016-01-11 17:52:26 (5912): Setting Network Configuration for NAT.
2016-01-11 17:52:26 (5912): Disabling USB Support for VM.
2016-01-11 17:52:26 (5912): Disabling COM Port Support for VM.
2016-01-11 17:52:26 (5912): Disabling LPT Port Support for VM.
2016-01-11 17:52:26 (5912): Disabling Audio Support for VM.
2016-01-11 17:52:26 (5912): Disabling Clipboard Support for VM.
2016-01-11 17:52:26 (5912): Disabling Drag and Drop Support for VM.
2016-01-11 17:52:26 (5912): Adding storage controller(s) to VM.
2016-01-11 17:52:26 (5912): Adding virtual ISO 9660 disk drive to VM. (vm_isocontext.iso)
2016-01-11 17:52:26 (5912): Adding VirtualBox Guest Additions to VM.
2016-01-11 17:52:26 (5912): Adding network bandwidth throttle group to VM. (Defaulting to 1024GB)
2016-01-11 17:52:26 (5912): Enabling shared directory for VM.
2016-01-11 17:52:26 (5912): Enabling scratch shared directory for VM.
2016-01-11 17:52:26 (5912): Starting VM. (boinc_00d1dfab3614e292, slot#4)
2016-01-11 17:57:26 (5912): Successfully started VM. (PID = '5912')
2016-01-11 17:57:26 (5912): Reporting VM Process ID to BOINC.
2016-01-11 18:02:27 (5912): VM is no longer is a running state. It is in 'poweroff'.
2016-01-11 18:02:32 (5912):
NOTE: VM failed to enter an online state within the timeout period.
This might be a temporary problem and so this job will be rescheduled for another time.


</stderr_txt>

Hi guys, could any of you try upgrading to Virtualbox 5.0.12 (link here) to see if that fixes it?

I check it. It´s works fine :)

Few combibation (Windows 7)

Boinc version 7.6.22 / 7.6.9 / 7.4.42 + virtualbox 4.3.12 = all works fine.
Boinc version 7.6.22 + virtualbox 5.0.12 = works fine too.

Boinck version 7.6.22 + virtualbox 5.0.10 = hypervision failed. This is Not c@h problem, vlhc and atlas@home crashed too.

Hmm, i believe that vbox 5.0.10 is problem in my Computer.

Rom Walton re-released BOINC 7.6.22 to include Virtualbox 5.0.12 which should fix this problem. If you're on Windows seeing these errors please download this newer version.

Thanks for the help everyone looking into this.

I've seen a VM problem you might want to watch for. Some of the older methods of detecting VM support will run only once per computer, so if that computer had it run before VM support was turned on in the BIOS or UEFI, that computer will never check for VM support again; instead, it will only remember the failure. Have you found a way to force a new check for VM support even while BOINC remembers a failure from a previous check?

Another problem on one of my computers - it is already supplied maximum memory compatible memory for the motherboard (8 GB), but VirtualBox requires so much memory (4 GB) that it cannot run more than one VM workunit at once, and even one leaves too little free memory to allow a reasonable selection of other BOINC projects to use the remaining CPU cores. Also, Windows Vista has rather poor memory usage for 32-bit applications (about as much memory required for the SYSWOW64 modules needed to allow 32-bit programs as for the programs themselves), so I can no longer use the console while your legacy application runs - it's still more memory hungry than nearly all applications for other BOINC projects I am interested in. Therefore, Cosmology@Home is set to No new tasks on that computer, and will stay that way unless you offer a way to select only applications that are less memory-hungry under Windows Vista. (When I get past my current reason for keeping Windows Vista, I expect to be more likely to replace the entire computer than to replace just the Windows version.)

I've seen a VM problem you might want to watch for. Some of the older methods of detecting VM support will run only once per computer, so if that computer had it run before VM support was turned on in the BIOS or UEFI, that computer will never check for VM support again; instead, it will only remember the failure. Have you found a way to force a new check for VM support even while BOINC remembers a failure from a previous check?

Yea this is definitely an issue, this is the best solution we've got at the moment. I've talked to Rom about better ways of checking for VM extensions and there may be a solution but its will likely be a little while till it can get implemented.

Another problem on one of my computers - it is already supplied maximum memory compatible memory for the motherboard (8 GB), but VirtualBox requires so much memory (4 GB) that it cannot run more than one VM workunit at once, and even one leaves too little free memory to allow a reasonable selection of other BOINC projects to use the remaining CPU cores. Also, Windows Vista has rather poor memory usage for 32-bit applications (about as much memory required for the SYSWOW64 modules needed to allow 32-bit programs as for the programs themselves), so I can no longer use the console while your legacy application runs - it's still more memory hungry than nearly all applications for other BOINC projects I am interested in. Therefore, Cosmology@Home is set to No new tasks on that computer, and will stay that way unless you offer a way to select only applications that are less memory-hungry under Windows Vista. (When I get past my current reason for keeping Windows Vista, I expect to be more likely to replace the entire computer than to replace just the Windows version.)

Sounds like two problems here.
1) For the legacy app its single threaded, so each job you run simultaneously will eat up a certain amount of memory, which can add up if you have a lot of cores and you run one job for each. Unfortunately not much to do there, and there's no plans to update the legacy app.
2) The new camb_boinc2docker is multithreaded so it doesn't suffer from this. In fact you only want one camb_boinc2docker job at a time, so maybe I'm not understanding why you say "cannot run more than one VM workunit at once" is a problem? That said, I'm quite surprised its using up so much memory for you, it should only be about 2GB of RAM max for a camb_boinc2docker job. If you do ever run more jobs, I'd be curious if you could check your task manager and let me know the names of the processes using the most RAM.

Wish I could report that VBox 5.0.12 has fixed the problem on my machine. No dice unfortunately.

http://www.cosmologyathome.org/result.php?resultid=36007392

---

Well VBox 5.0.12 worked better than 5.0.10 (that is to say I was able to complete maybe two out of three work units).

However, going back to VBox 4.3.12 worked best of all - I haven't had a single failure so far and it works perfectly well with BOINC 7.6.22.

OK so it seems 5.0.12 didn't really fix it after all. Thanks for the reports. The next step is we've got to update vboxwrapper a bit to give us a better error report in the case of these failures, we'll try to do that soon, then we can see what's actually going on.

I just had 14 tasks fail (I finally aborted them) with: "task postponed 86400.000000 sec: Communication with VM Hypervisor failed."

This is one that's actually updated to 5.0.16 virtualbox, so I know it's not a latest version issue.

It's host 280861 and there are plenty of tasks to view the stderr output from.

---

I've had the same problem with the same version of the VM. Mine started after a failed installation of an FEM package who's name escapes me, which used Docker to set up a virtual machine. Not sure if this was the cause after reading this. I'm running on an AMD 8 core FX8350 with 32Gbytes of RAM, GTX 950 video card, with a Tesla M2090 GPU so I don't think this is a memory related problem. I also have similar errors to what you have all listed and like you it is just ATLAS and Cosmology@Home that are suffering.

---

Unfortunately another 20 Planck tasks exhibited the same behavior after completely restarting the (physical) machine.

For now I've detached from the project on this machine. I'll try on Monday to re-install VirtualBox and attach to the project again.

---

Hi,
since yesterday on one of my machines I'm experiencing this problem of VM Hypervisor failed to enter an online state in a timely fashion.
I'm crunching since three weeks and this is the first time it happens, I've already detached and reattached to the project but no improvements so far.
Please help.
Many thanks