Can't run more than 63 guests

[drachev]

Hardware - MB P7H55-M/USB3, CPU I7-870, RAM 4х4G DD3-1600, HD 500G WD Caviar Black, SSD Crucial 256G.
Host - W7 Enterprise, VirtualBox of latest version.
Guests - Win XP, have tried to use many different builds of mini XP (Lite, Micro etc) as well as licensed XP with the same result.

All is fine while running till 63 guests.
Problems arise after starting just 64 vm (or more).
After a seemingly successful shutting down any of running machine
(regardless of by the command line or via GUI)
VB Manager shows this mashine in "Powered Off" state,
!! BUT !! BUT!! The "Settings" and "Start" buttons are disabled
and its process is still active in Task Manager!
So I need to kill the corresponding process in Task Manager manually. After killing, Task Manager shows "Powered Off" state correctly .
While running 63 VMs there are about 1-1,5GB free RAM, CPU usage about 20-40%, not any significant delay at HDD or SSD drives,
i.e. there are enough recourses to run for at least an additional 10 VM.
I was really successful in starting up to 80 VMs, but unfortunately with the described above glitches.

What could be the problem and the reason of so strange limitation to 63 VMs?

[mpack]

How many CPU cores does your host have? How much RAM does the host have, and how much are you giving to each VM?

[Grumlin]

so, I have the same problem too
I use win 7 x64 Ultimate, virtualbox the latest version too
guest - XP
I run 112 VMs, but the problem the same, after 63 VMs I have that describe above
Please, I want to use VirtualBox and don't want to use VMware
my configuration Core i7-3820 LGA 2011 4 core - 8 threads and I have 32 gb memory, for each I give 160 mb memory, but I have this problem even when I gave 192 mb each virtual machine

By the way, I use HDD and SSD too, but I don't think that the problem with it
and when I run 64 VM, the process of VBoxSVC.exe begin load processor more than before(before 1-2%, but after 64 vm 13-15%(1 thread)).

that problem on all platforms, because on 2 comp I have problem too(AMD FX-4100, 4 cores 4 threads)

[drachev]

How many CPU cores does your host have? How much RAM does the host have, and how much are you giving to each VM?

CPU I7-870 has 4 cores - 8 threads
RAM 4х4G=16G
I've tried a different amounts of memory from 128M to 64M per guest, but the digit 64 remains the magic number causing the problem described above.

[Technologov]

There are known users, that successfully run over 1,000 VMs on Linux hosts. (with 1 TB of RAM and 128 CPU cores)

See the VirtualBox limits

What could be the reason for your issues ?
Perhaps internal limitations of VirtualBox-on-Windows. (either at service level, VBoxSVC, or the kernel driver level, vboxdrv)

Worth sending an email to mailing-list. If no response in 3 days, you can open a bug.

[drachev]

There are known users, that successfully run over 1,000 VMs on Linux hosts. (with 1 TB of RAM and 128 CPU cores)

See the VirtualBox limits

What could be the reason for your issues ?
Perhaps internal limitations of VirtualBox-on-Windows. (either at service level, VBoxSVC, or the kernel driver level, vboxdrv)

Worth sending an email to mailing-list. If no response in 3 days, you can open a bug.

Thanks for reply!
I know there isn't such limitation on Linux hosts as well on Windows Servers one, but should I change my OS due to the bug in VB?
Ok, will send an email and so on.

[klaus]

This limit to 63 guests is a known issue on Windows. MS in their infinite wisdom limited the array of objects one can wait for to 64 (see the zillion of questions in the Web about how to overcome this limit of WaitForMultipleObjects). It took them until Vista to introduce an API which can handle more (without creating lots of almost dummy threads), but many users are still on XP so this is not an option.

[Grumlin]

Did you get a response about this problem? if no, will you open a bag?

[drachev]

It looks like this limitation is really based on windows WaitForMultipleObjects limit.
I'm not sure - whether Oracle can fix such a windows error, so does it make sense to open a bug?

[Grumlin]

It looks like this limitation is really based on windows WaitForMultipleObjects limit.
I'm not sure - whether Oracle can fix such a windows error, so does it make sense to open a bug?

Oh, it's very bad... but maybe need to try

[Martin]

If you look through this forum you will recognize that "klaus" is one of the VBox developers from Oracle. Therefore I think that opening a bug will get you a similar answer like his response here in this thread.

[Grumlin]

so, but I don't understand, why the developer didn't so, that the process of VBoxSVC didn't load full thread???
or that on each 63 Machines starts new VBoxSVC(VBoxSVC, VBoxSVC1, ...)?