VBoxSVC Hangs when cycling many Headless VMs

[dsterne]

I am currently running 4.3.36r105129 on CentOS 6.3 (2.6.32-279.11.1.el6.x86_64), and I am constantly cycling about 120 short-lived (~60 seconds) headless linked clones (all linked to the same base VM). After running for some time, and it seems pretty random, all VBoxManage commands begin to hang (no return).

I've attached an strace -f -p to the VBoxSVC process during a hang (PID 665541 in this case), and I am seeing:

[pid 665541] futex(0xa0a764, FUTEX_WAIT_PRIVATE, 1, NULL <unfinished ...>
[pid 666826] <... restart_syscall resumed> ) = -1 ETIMEDOUT (Connection timed out)
[pid 666826] futex(0x7ffe80025170, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 666826] futex(0x7ffe80025144, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 700047, {1464018266, 342966794}, ffffffff) = -1 ETIMEDOUT (Connection timed out)
[pid 666826] futex(0x7ffe80025170, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 666826] futex(0x7ffe80025144, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 700049, {1464018267, 341676697}, ffffffff) = -1 ETIMEDOUT (Connection timed out)
[pid 666826] futex(0x7ffe80025170, FUTEX_WAKE_PRIVATE, 1) = 0

The PID of 666826 no longer exists in the process table. From what I have seen, there is no specific VM or clone which is causing the issues. In order to get things flowing again, I need to pkill -6 VBoxSVC.

[Frank Mehnert]

Please reopen if still relevant with VBox 5.1.22.