QA » openQA Project

observation¶

svirt based tests can die when booting before installation because the image is unusable, e.g. see
http://lord.arch/tests/1048/file/autoinst-log.txt

Domain openQA-SUT-12 defined from /var/lib/libvirt/images/openQA-SUT-12.xml


20:13:09.4589 Command's stderr:

20:13:10.7595 Command's stdout:


20:13:10.7596 Command's stderr:
error: Failed to start domain openQA-SUT-12
error: internal error: process exited while connecting to monitor: 2016-06-13T20:17:10.483081Z qemu-system-s390x: -drive file=/var/lib/libvirt/images/openQA-SUT-12.img,if=none,id=drive-virtio-disk0,format=qcow2: qcow2: Image is corrupt; cannot be opened read/write


20:13:10.8867 # Test died:
{
  'args' => [],
  'console' => 'svirt',
  'function' => 'define_and_start'
}
virsh start failed at /local/os-autoinst/consoles/sshVirtsh.pm line 392.

also seen in
http://lord.arch/tests/1055
http://lord.arch/tests/1087

steps to reproduce¶

Run zkvm tests often, happens in about 3/20 runs, at least in my setup.

problem¶

H1. specific to my openQA+reserved_s390x@zkvm worker (VIRSH_HOSTNAME=s390pb.suse.de, VIRSH_GUEST=10.161.145.7, VIRSH_INSTANCE=12)
H2. the s390x guest is "reused" or the image is cleanup up by someone else
H3. out of disk space and return value of qemu-img command is not properly checked?

suggestion¶

We could improve the check for the image being present before we try to access it again after reboot.

workaround¶

sporadic, so retrigger