Bit of context. Few VM's running on KVM (SmartOS), using kernel 2.6.32.7 have been rebooting. Nothing in the Qemu/KVM logs to suggest it's a Qemu issue.
Unfortunately while testing kdump i discovered a problem that causes the kdump kernel to panic when writing the dump disk in the virtio blk driver.
So I decided to just set up the kernel to log to a serial device to capture a stack trace. Had a machine reboot and nothing other than boot sequence, no stack trace. Early printk is set in the kernel configuration as well.
While these VMs will just be blown away and re-provisioned using a newer image it has raised a question.
Namely why is the VM rebooting in the first place, Qemu is not exiting and been started by an external process the guest is clearly rebooting (uncleanly due to the state of ext on boot).
However AFAIK it should not reboot, wether panic, oops, soft lockup or hardlockup, it should remain up (even if lockedup) kernel.panic systcl is set to 0.
kernel.panic = 0
kernel.panic_on_oops = 0
kernel.unknown_nmi_panic = 0
kernel.panic_on_unrecovered_nmi = 0
kernel.panic_on_io_nmi = 0
kernel.softlockup_panic = 0
kernel.hung_task_panic = 0
vm.panic_on_oom = 0
What else would cause the linux kernel to decide to reboot, or am i misunderstanding any of the above sysctls.
0 Answers