Development for LHC@home

This new version is for Windows only. It provides a rebuilt vboxwrapper. With the previous vboxwrappers, the Windows version differs from the Linux and Mac versions. Whereas the Linux and Mac versions use VBoxManage to control the VMs, the Windows version uses the API. The main issue is the the behavior differs depending on which OS is being used. The other issue is that updates to the vboxwrapper are urgently needed when the API is changed during a Vbox update. Lucky, as the code was written using ifdef WIN32, it was trivial to rebuild the vboxwrapper so that VBoxManage is used for Windows rather than the API. Please let us know if you see any issues with this build.

Not an issue with the new wrapper, but I saw already during some runs the growth of the Benchmark vdi-file in the slot folder.
Did not know what the disk-bound was, but then I got the answer:

lhcathome-dev 27 Feb 16:46:34 Aborting task Benchmark_3570_1488026643.918139_0: exceeded disk limit: 7806.10MB > 7629.39MB

It seems the disk_bound is set to 8000000000 bytes and this time the vdi-file was exceeding that value.

Hi Laurence,

I discovered that the vm_image.vdi is removed from the slot directory, but not from VirtualBox Media Manager.
In Virtual Media Manager the disk is marked with an exclamation mark.

UUID: c0f46db0-1029-40b8-9c1d-8a961537b484
Parent UUID: base
State: inaccessible
Access Error: Could not open the medium 'D:\Boinc1\slots\0\vm_image.vdi'.
VD: error VERR_FILE_NOT_FOUND opening image file 'D:\Boinc1\slots\0\vm_image.vdi
' (VERR_FILE_NOT_FOUND)
Type: normal (base)
Location: D:\Boinc1\slots\0\vm_image.vdi
Storage format: VDI
Format variant: dynamic default
Capacity: 20480 MBytes
Size on disk: 4558 MBytes
Encryption: disabled

The cleanup after the task finished should be improved.

Maybe it's simple and you only have to shuffle the order to cleanup the VM.

Your wrapper now unregisters the VM first and then closes the hdd medium.

Maybe it's simple and you only have to shuffle the order to cleanup the VM.

Your wrapper now unregisters the VM first and then closes the hdd medium.

I manually replayed the cleanup procedure in the sequence used by the wrapper and could not reproduce the left behind vdi in VBox Media Manager.

With the new wrapper it's also not happening always (sigh), so maybe the commands for the cleanup are (sometimes) too fast after each other.

I spotted earlier that there were some Benchmark tasks available so I thought I'd give them another try, having had mixed success in the past.

1 became Unmanageable and reported as Not Started by Deadline on a Boinc restart.
1 lasted only 94 seconds but was valid.
2 ran to completion in 800 seconds CPU time.
All cleaned out their respective slots but 1 of the 800s ones left behind a ghost .vdi in Vbox Virtual Media Manager which had to be manually removed.

Anyway, real reason for the post is that all these tasks had a 1 hour deadline!. I don't really expect that this was by design and that there must have been a typo in the release of these tasks.

Another pair has just started so I'll keep an eye on them too.
Both became Unmanageable and reported soon after a Boinc restart to get them going again. All tidied up, no debris.

One more pair then off to bed.
No issues.
All tidied up, no debris.

Back to Sixtrack resends overnight.

Thanks. Were were using this application to try out a new version of the wrapper. It will be fixed soon. the 1 hour deadline is probably correct as the task should only run for about 5 minutes.

All 6 of this evening's tasks ran fine, cleared their slots, validated and credited. However, all 6 left an empty, unattached, yellow triangle, ghost image reference in Vbox Virtual Media Manger.
From a sample stderr:
2017-05-22 20:03:53 (5932): Powering off VM.
2017-05-22 20:03:56 (5932): Successfully stopped VM.
2017-05-22 20:03:56 (5932): Deregistering VM. (boinc_77bf12a892cf81af, slot#4)
2017-05-22 20:03:56 (5932): Removing network bandwidth throttle group from VM.
2017-05-22 20:03:56 (5932): Removing storage controller(s) from VM.
2017-05-22 20:03:57 (5932): Removing VM from VirtualBox.
2017-05-22 20:03:57 (5932): Removing virtual disk drive from VirtualBox.
20:04:03 (5932): called boinc_finish(0)

It reports the highlighted line but doesn't seem to actually do it successfully. The VM itself gets removed but the reference remains until manually removed.
Don't know if that's a Boinc/Vbox miscommunication or the wrapper?

Boinc 7.6.33, Vbox 5.1.22, 26198ab1 wrapper, Win10.

Current 2 have just become Unmanageable. 1 shows in Vbox as Aborted, the other as Powered Off. A couple of LHC Theory tasks are due to finish within the hour so I'll not restart Boinc until then in case those get upset. The Benchmark tasks will be past deadline by then. I'd rather lose their combined 30mins than 28 hours.

I'm currently downloading another pair on a different host, just for comparison, but the deadline being so short means they are likely to timeout before the 849MB image finishes downloading.

We are actively working on the wrapper, just released version 0.12.