Zhengang Li wrote:
zli@linux.vnet.ibm.com wrote:
diff -r 5c77329cb53e -r 277b821188f2 suites/libvirt-cim/lib/XenKvmLib/const.py --- a/suites/libvirt-cim/lib/XenKvmLib/const.py Fri May 30 14:26:38 2008 +0800 +++ b/suites/libvirt-cim/lib/XenKvmLib/const.py Wed Jun 04 10:06:11 2008 +0800 @@ -85,3 +85,5 @@ #vxml.LXCXML LXC_init_path = os.path.join(_image_dir, 'cimtest_lxc_init') LXC_default_tty = '/dev/ptmx' +LXC_default_mp = '/tmp' +LXC_default_source = '/tmp/lxc_files'
The LXC_init_path used to be created by 'create_lxc_files' in vxml.py. It's not created here. But somehow libvirt shows me a running container even if the init_path file doesn't exist.
I believe this is a bug in the containers support in libvirt. The container attempts to execute the script, but the execution fails. The ID of the container is the pid of the process the container executes to run the script. If you no longer see the pid in the ps output, then the container is not actually running.
The same with LXC_default_source. It's a non-exist directory.
I am inclined to move the init_path and default_source to the place where xen/kvm kernel image and disk image is created. Then we don't need to create a init_path upon every test case execution.
Thoughts?
Initially, I thought it'd be a bad idea to leave the script lying around. However, all the script does is run bash, so I don't think it will be a problem to live it lying around in /tmp. -- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com