Co jakiś czas jedna z wirtualek przestawała działać. Problemem okazał się KSM. Możliwe, że jakiś bug. Rozwiązanie problemu jest banalne. Wystarczy wyłączyć KSM.
systemctl disable ksmtuned
Poniżej wycinek z logów wirtualki.
Jan 29 10:21:20 plus kernel: kworker/0:0: page allocation failure: order:0, mode:0x310da
Jan 29 10:21:20 plus kernel: CPU: 0 PID: 25322 Comm: kworker/0:0 Not tainted 3.10.0-957.1.3.el7.x86_64 #1
Jan 29 10:21:20 plus kernel: Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.10.2-0-g5f4c7b1-prebuilt.qemu-project.org 04/01/2014
Jan 29 10:21:20 plus kernel: Workqueue: events_freezable update_balloon_size_func [virtio_balloon]
Jan 29 10:21:20 plus kernel: Call Trace:
Jan 29 10:21:20 plus kernel: [<ffffffffb8361e41>] dump_stack+0x19/0x1b
Jan 29 10:21:20 plus kernel: [<ffffffffb7dbcaa0>] warn_alloc_failed+0x110/0x180
Jan 29 10:21:20 plus kernel: [<ffffffffb835d44e>] __alloc_pages_slowpath+0x6b6/0x724
Jan 29 10:21:20 plus kernel: [<ffffffffb7dc1105>] __alloc_pages_nodemask+0x405/0x420
Jan 29 10:21:20 plus kernel: [<ffffffffb7e0df68>] alloc_pages_current+0x98/0x110
Jan 29 10:21:20 plus kernel: [<ffffffffb7e3dc55>] balloon_page_alloc+0x15/0x20
Jan 29 10:21:20 plus kernel: [<ffffffffc0475811>] update_balloon_size_func+0xb1/0x290 [virtio_balloon]
Jan 29 10:21:20 plus kernel: [<ffffffffb7cb9d4f>] process_one_work+0x17f/0x440
Jan 29 10:21:20 plus kernel: [<ffffffffb7cbade6>] worker_thread+0x126/0x3c0
Jan 29 10:21:20 plus kernel: [<ffffffffb7cbacc0>] ? manage_workers.isra.25+0x2a0/0x2a0
Jan 29 10:21:20 plus kernel: [<ffffffffb7cc1c31>] kthread+0xd1/0xe0
Jan 29 10:21:20 plus kernel: [<ffffffffb7cc1b60>] ? insert_kthread_work+0x40/0x40
Jan 29 10:21:20 plus kernel: [<ffffffffb8374c37>] ret_from_fork_nospec_begin+0x21/0x21
Jan 29 10:21:20 plus kernel: [<ffffffffb7cc1b60>] ? insert_kthread_work+0x40/0x40