virtualbox.org

Host: Windows 10 Pro 64 bit (version 1607 build 14393.693), Intel i7 (8 core), 16Gb Memory
Guest VM: Linux 7.1 64 bit (running Oracle 12c)
VirtualBox: 5.1.14 with Oracle_VM_Guest_additions 5.1.14

Problem

I have been runinng a Virtual Box environment for sometime with no issues. All existing guest VMs now lockup when started (3 linux guest vms with slightly different states of oracle database). Starting a VM from the Virtualbox Manager ususally takes less than 1 minute. A startup of a vm guest now takes > 10 mins and eventually locks up. Guest O/S boots (Linux server boot messages appear in guest window) but boot is very slow, guest system eventually locks up and displays error message on guest vm window:-

"Oh No Something has gone wrong
A problem has occurred and the system cant recover
Please log out and try again "

Unable to log out, vm completely frozen and have to user VirtualBox Manager to manually powerdown.

Guest storage set up on Samsung SSD 500gb drive, set to run with 4 cores (out of , 8192mb mempry, 128mb video memory. 2d & 3D turned off. Oracle Guest Additions set up.

At the point of lockup, windows Task Manager (on the host) shows 50% cpu usage for the virtualbox manager process, total cpu usage < 75% so not maxed out on cpu. Memory for VM on Task Manager shows 115mb, total memory in use < 25%. However disk i/o is very low, 0.1 MB/s - which is low (lock up at disk i/o level?).

I have checked ssd drive (used Samsung Magician) - no problems found. Drivers are up to date.
Permissions at windows level on ssd drive and directories are not restricted.
I have run Dell hardware diagnostics to check memory/bus/cpu etc - no problems.
There are no errors in the Windows event logs.
Virus scanning is turned off on the drive/directory that holds vms (Norton 360)
Hyper-V is disabled and not running.


Attempted fixes and tests.

1. Reverted to 5.0.30 by reinstalling (reset guest additions to 5.0.30 in Manager Preferences). Same problem on all guest vms.
2. Reset guest vm to use 1 cpu and then through to 5 cpus. Same problem.
3. Deinstalled VirtualBox completely (deleted .Virtualbox directory) and installed 5.0.30 from scratch. Copied VM machines from backup from earlier 5.0.30 installation timeframe (guest vm working without problem in this backup timeframe). Same problem.
4. Deinstalled VirtualBox completely (deleted .Virtualbox directory) and installed 5.1.14 from scratch. Rebooted PC and stopped all non essential processes (outlook, business skype, backup, etc). Changed default storage location to non ssd drive, standard sata drive. Imported Oracle Developer Day Appliance to a non SSD drive (used appliance as a base line test to establish if issue was with generic or VM specific). Started with no problems, but did not attempt any major read /write tasks.
5. Copied backup of most recent guest VM to non ssd drive and added VM in VirtualBox Manager. Unchecked SSD option for disks 1 & 2 under Settings - Storage. Copy of VM "backup" started OK (upgraded guest extensions in Linux guest once started). Task Manager now showing lower cpu usage (30% max when using Oracle development tools in guest) and disk i/o is much higher (15 - 20 MB/s during startup, 5 - 8MB when using Oracle development tools in guest).
6. Re enabled /started other windows task (Outlook etc). No problems, vm continues to be useable and does not lock up.
7. Cloned working copy of VM from step 5 above to SSD drive. Checked (set to on) SSD option for disks 1 & 2 under Settings - Storage. Problem now re occurs.
8. Recreated step 5 and used VM (on sata drive) for a few hours. After a few hours (and large saves/write to database on VM guest), guest now locks up. Task Manager shows i/o as neglible even though large write being attempted on guest. Anti virus process is not showing as having been evoked (ie Norton 360 running at very low % in background and does not appear to be scanning write). CPU for Virtual Box at 50%. Guest manually powered down via Virtual Box Manager.
9. Tried to start Oracle Developer Day Appliance installed at step 4. This now also won't start and lockups with "Oops" message.
10. Rebooted workstation. Oracle Developer Day Appliance now starts and runs OK. Shutdown Oracle Developer Day Appliance cleanly within guest o/s. Then started VM that froze at step 8. No problem starting VM, i/o in Task Manager > 20 MBs. Linus guest bug reporting tool not showing any errors other than Gnome Setting Monitor killed by sigdev (which woudl have happened at power down).


This seems to suggests that there is a problem of some sort with the o/s on the workstation, particularly with ssd drive. There are no problems an a Windows level with the ssd drive, large 80gb files can be copied and written to/from ssd drive with no problems. However, as the VMs have been working previosly without problem, I suspect that the conflict has arisen post a Windows 10 update.

The workstation has been automatically upgraded over time with Windows 10, Chrome/Firefox etc, but no other changes to the host environment. There are no other issues with other software on the workstation. The virtualbox environment and the Linux VMs has been running successfully for nearly two years, with periodic upgrades of Virtualbox.

The logs files are attached (slightly renamed so I can track which vm log files came from). I am not a systems guy (expertise is in Oracle), so not sure how to interpret log files to see if any obvious issue. The zip files are named as below:-

a) VBoxLogSata2. Log files for successfull power up and down of backup at step 5
b) VBox-vmssd1. Log files for VM on ssd drive as at step 7 above.
c) VBoxSataLockup. Logs files after running VM guest for awhile per step 8.

Input, advise and help (desperately and) very gratefully received.

Thanks

Paul

Note to moderator: I originally posted a (shorter) version of this problem as a Linux guest issue, but I suspect the problem is at the windows level (i/o issue with windows o/s?) and that therefore this should be in the Windows Host forum. There have been no responses on the Linux thread but I will update the Linux thread appropriately if a resolution is found.

Lots of possibly useful info there except for one thing:

"Oh No Something has gone wrong
A problem has occurred and the system cant recover
Please log out and try again "

This needs to be cleared up. What is the exact message? Is it a Virtualbox popup? A host BSOD?
Sorry this is a guest message as mentioned:

guest system eventually locks up and displays error message on guest vm window

The Intel Core i7 Q820 is a 4 core CPU with 8 threads, but for virtualization only physical cores really matter.
So if you see 50% CPU in Windows task manager your computer can ba already running at full load on all available cores.
If you want best performace on such a system don't assign more than 3 cores to a guest to leave one for the host to handle the virtual machine.

Thanks for the quick response.

The message appears in the Linux guest windows. It appears after the linux boot messages are displayed and is generated by the guest o/s, not the virtual box application. My guess would be that the guest o/s is unable to write data to the host o/s and has therefore locked up when trying to spawn it's usual start up processes. As the guest o/s has locked up, it is hard to determine exactly what process is locking up.

Regards

Paul

Thanks Martin for looking at the logs. I will amend the cpu count as you recommend.

However, one of the test I did was to to alter settings to be at 1 through to 5 cpus. This made no difference to the occurence of the problem. The VMs have also been (incorrectly) set at 4 cpus for the last 18 months with no problem, so I suspect the issue is elsewhere.

Regards

Paul

@Martin I should have added that the tests with the Oracle Developers Appliance was set at 2 cpus. This ran OK on the Sata drive but when cloned to the ssd drive, it had the same problem. However, no write load was attempted in the guest to test the Sata environment.

Sorry, I didn't have much time earlier.
It was just a quick general comment about your system and VM sizes, not really directly relevant to your problem.