The machine xxx does not have an open session

[ToddAndMargo]

Hi All,

host: CentOS 5.5 x64 guest: windows server 2003 R2, SP2 VB: 3.2.10

The guest is working fine, but when I try to control it, I get:

/usr/bin/VBoxManage --nologo guestcontrol execute VB-PPro2 "C:\NtUtil\set.bat" --arguments "a b c" --username "winadmin" --password "notellin" --wait-for stdout

ERROR: The machine 'VB-PPro2' does not have an open session Details: code VBOX_E_INVALID_SESSION_STATE (0x80bb000b), component Machine, interface IMachine, callee nsISupports Context: "OpenExistingSession(a->session, uuid)" at line 297 of file VBoxManageGuestCtrl.cpp

Also: /usr/bin/VBoxManage --nologo list runningvms Does not show the guest.

Work around: completely power off the guest (restart does not work) and the start the guest back up from the powered off state.

Please fix. :-)

Many thanks, -T

[Klaus Espenlaub]

We need lots more information - from the symptoms it appears as if it's a VBoxSVC crash or other API communication problem. How did you start the VM? Do other commands to control the VM work? Can you check if there's a core dump? If core dumps are disabled you need to enable them...

[ToddAndMargo]

Replying to klaus:

How did you start the VM?

/usr/bin/VBoxHeadless --startvm $VM_Name & > $Log 2>&1

Do other commands to control the VM work?

No. ("does not have an open session" errors)

Can you check if there's a core dump?

Where?

If core dumps are disabled you need to enable them...

How?

Next time one crashes, is there some test you would like me to perform?

Many thanks, -T

[ToddAndMargo]

I can VRDP into the guest without problems.

[Klaus Espenlaub]

The core dumps I'm expecting are created in the current directory of the VBoxSVC process. Normally that's the home directory of the user.

​https://www.centos.org/modules/newbb/viewtopic.php?topic_id=29032&forum=38 points to /etc/profile for enabling core dumps - we can't know answers for all distros...

It's expected behavior with VirtualBox 3.2 that the VBoxHeadless processes continue running (with working VRDP) when VBoxSVC died. With VirtualBox 4.0 this will change, the VM will power itself down (with an appropriate line in VBox.log), as such zombie VMs are uncontrollable via the API, as you already noticed.

[ToddAndMargo]

Not finding a core dump. And, everything is well behaved today.

Did upgrade to 3.2.12 last week.

Any specific tests you'd like me to do or just wait for the next core dump?

-T

[Klaus Espenlaub]

Since there is no hint yet what leads to the crashes I can't suggest tests.

If you have VMs running all the time it would be worth remembering the process ID of VBoxSVC, because in such a case it must not change. If it does change there was either no VirtualBox client running at one time (triggering automatic VBoxSVC termination after 5 seconds), or it did crash.

[ToddAndMargo]

I have a machine running right now that does not have an open session. No sign of a core dump. Any test you'd like me to try?

Also, now that 4.0.x is out (and unusably slow for us), is patching on 3.2.12 a dead issue?

-T

[Frank Mehnert]

Still unknown if there will be ever a public 3.2.14. Which 4.0 version is slow for you, does the slowness start with 4.0 or later? And which workload did you use to verify that 4.0.x is slower than 3.2.12?

[ToddAndMargo]

Replying to frank:

Still unknown if there will be ever a public 3.2.14.

Bummer

Which 4.0 version is slow for you, does the slowness start with 4.0 or later?

4.0.2

And which workload did you use to verify that 4.0.x is slower than 3.2.12?

Windows XP Guest. All functions were at least twice as slow.

You should look at this benchmark: ​http://www.phoronix.com/scan.php?page=article&item=linux_kvm_virtualbox4&num=1 Does not show 4 as twice as slow (what I encountered), but it does show it as significantly slower.

[Frank Mehnert]

Sorry for asking again: Did you ever test VBox 4.0 so are you 100% sure this is a 4.0.2 regression?

[ToddAndMargo]

Hi Frank,

Only tested 4.0.2; never tested any other 4.x. Since then, I have been to busy to test 4.0.4. I checked the change logs and found nothing relating, so I was in no hurry.

Right now on a production level server (3.2.12), I am getting tons of iaStor error in the Windows system log and a data corruption in a Fox Pro database, so I am a bit tied up. I will be changing our the IATA89ENU v8.9.0.1023 driver for a straight IDE controller to see it that helps. There is no rest.

-T