A running VM is listed as Powered Off

[mlw]

I have a VM that has been running for 11 days, but VBoxManage showvminfo and the VirtualBox Manager window both report that is it powered off.

The machine is working perfectly AFAICT (its console is alive and processes are running in it).

# ps -wwo lstart,command 6686

STARTED COMMAND

Wed Jul 6 12:45:57 2011 /usr/lib/virtualbox/VirtualBox --comment barker --startvm 91b05e8b-b485-49cb-849c-eb5dfed6dcf6 --no-startvm-errormsgbox

# VBoxManage showvminfo barker | grep State State: powered off (since 2011-07-06T10:44:00.000000000)

The log contains thousands of messages like this:

199:10:30.740 ERROR [COM]: aRC=NS_ERROR_FAILURE (0x80004005) aIID={12f4dcdb-12b2-4ec1-b7cd-ddd9f6c5bf4d} aComponent={Session} aText={Failed to query the session machine (NS_ERROR_CALL_FAILED)}, preserve=false

[mlw]

compressed log (uncompressed is 13MB)

[Ron Gould]

Identical problem and symptoms.

VirtualBox Host:
- CentOS 5.7 64-bit
- VirtualBox version 4.1.2 r73507

Virtual Machines
- CentOS 5.7
- Debian 6
- Ubuntu 11.04

Problem occurs on Host servers with Xeon, i3, and i7 CPUs.

VMs were all built and started using VirtualBox GUI interface.

Examples:

vbox@host01:~$ VBoxManage showvminfo "deb64_wiki01'
...
State:           powered off (since 2011-09-30T04:21:23.000000000)
...
root@wiki01:~# uptime
 16:28:55 up 1 day, 18:58,  2 users,  load average: 0.00, 0.00, 0.00


vbox@host01:~$ VBoxManage showvminfo "cent53_db"
...
State:           aborted (since 2011-09-29T01:20:45.000000000)
...
[root@db ~]# uptime
 16:32:15 up 2 days, 20:22,  2 users,  load average: 0.00, 0.00, 0.00


vbox@host01:~$  VBoxManage list runningvms
vbox@host01:~$ 

vbox@host01:~$  VBoxManage list vms
"cent53_db" {b279b9f8-9f59-4188-a8c4-ecf7ca07a500}
"cent57_baseline" {80b85e42-7386-4966-9002-615bcd2d0eaa}
"cent57_vanilla" {50b40551-946f-48fb-872c-4bddec8acc45}
"cent57_nilmon" {437d9eb8-6043-4909-b548-271a0cddbe05}
"cent57_amimon" {45216538-7132-4742-98c0-8bc1a37ebf93}
"deb64_baseline" {684dcf2f-a466-4e87-a1a5-17fd786833e3}
"ubi32_wiki" {ce795256-3ab0-4dea-8cb9-403dca217e09}
"deb32_wiki" {738c47f0-651a-451e-9bec-507a37fe3ce9}
"deb64_wiki01" {070e7a15-e07d-43a4-837b-a5e1688ad65c}
vbox@host01:~$

[Ron Gould]

Samwise here again.

Tracked the problem down to a location between the chair and keyboard.

Basically VBoxManage, the VBox console, and vboxtool all report accurate data, PROVIDED THAT the same user session is used to start the VMs.

If I start a session using the VBox console, then ssh in -- even as the same user -- VBoxManage et al will be confused. Or if you are using VNC carelessly; same issue.

To determine if this is the issue, simply run "users" on the host. If you see your name multiple times, one of the other logins actually knows about the running instances.

However, I have now moved to strict use of a single user and vboxtool to start VMs, and either use sudo poweroff from WITHIN the VM or vboxtool stop vmname. All under the same user, with clean logouts OR the use of screen.

After moving to this model, I can report 100% of the problems are eliminated, and I have never had a VM go missing again.

I am happy to provide details to whomever would like them.

• Sam :D

Replying to samwisekoi:

Identical problem and symptoms.

VirtualBox Host:
- CentOS 5.7 64-bit
- VirtualBox version 4.1.2 r73507

Virtual Machines
- CentOS 5.7
- Debian 6
- Ubuntu 11.04

<skip>

[arb]

I'm not sure I would be so quick to place the problem with the user. I have experienced the same symptoms and would say the problem is with VirtualBox. As long as I login again to the same user account I should see the VMs which I started, it's as simple as that. If I don't then VirtualBox is broken. I am currently experiencing this problem so if the developers would like my assistance to troubleshoot it then I can help right now.

[mark1ewis]

This ticket may possibly be related to #12250

[repadelft]

I am having the same trouble, running VirtualBox (currently 4.3.8) under Fedora. It wasn't always there, though.

I set up a Windows XP virtualbox to print for me, to a printer that is windows only. I created a cups script that checks whether the VM is running, all using runuser, starts the VM if not, with "VBoxManage startvm 'Windows XP' --type headless", pushes a print file there and starts a print command, through gsview.

When I initially set it up, a few years ago, (don't remember exactly when) this would be clearly visible; the VirtualBox graphical interface would show that the virtual machine was started, then stopped again.

Now that no longer works, and if I am not careful i.e. running the virtual machine and then printing, the VM is started twice and corruption is the result. Long live snapshots!

Somehow the mechanism through which VirtualBox communicates with the VM's might have been changed, either through the OS or through VirtualBox itself.

Are there any options that can be used to connect to a virtualbox started in another session?

[Frank Mehnert]

The communication mechanism didn't change. If you start a VM and the VirtualBox GUI does not list this VM as started then this usually means that the VBoxSVC service crashed for some unknown reason. So please check if that's the case. If so, you can probably provide a core dump?