Since the stable-compute-uuid nova feature the compute nodes created via
VIRT_DRIVER=fake
cannot be restarted as these computes are not writing the compute_id
file to disk at first startup. Therefore any subsequent restart will
fail as nova-compute will refuse to start due to the missing compute_id
file but having a service already in the DB.
After this patch fake-virt uses a variant of the fake virt driver that
actually writes compute_id file to disk. To allow multiple fake computes
running on the same machine each compute now has a separate state_path
created so each can store a separate compute_id file.
Signed-off-by: Balazs Gibizer <gibi@redhat.com>
Change-Id: I813cab3c89554d1e319b2b3e5c3affd8ec5d887e