DB> This is a long overdue followup to my previous set of patches to
DB> make the LXC driver use the new domain XML apis.
Have you been able to test this with NET_NS support enabled yet?
I am hitting the same issue I was before. On starting an LXC domain
with a network interface, the daemon finishes the action and reports
a good status to virsh. The domain fails to start, with the following
in the per-domain log file:
DEBUG: lxc_container.c: lxcContainerStart (clone() returned, 8082)
libvir: Linux Container error : internal error read of fd 6 failed: Input/output error
DEBUG: veth.c: vethDelete (veth: veth2)
However, now, the daemon crashes well after leaving the LXC driver's
domain startup process. From the timing, I'd guess it's in some event
(or SIGCHLD) handler, but I can't reproduce it in gdb to get more
information.
This looks to be identical in root cause to what I was seeing with the
previous set of patches, and was unable to figure out why file
descriptor 6 was getting prematurely closed. I'll get started on
debugging it again tomorrow, but it might be good if you can reproduce
it as well :)
--
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms(a)us.ibm.com