recent change to qemu.conf breaks some deployments
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
kolla |
Fix Released
|
Critical
|
Steven Dake | ||
Liberty |
Fix Released
|
Critical
|
Steven Dake | ||
Mitaka |
Fix Released
|
Critical
|
Steven Dake |
Bug Description
The Dockerfile template for the nova_libvirt container has a sed one-liner intended to set the parameter "stdio_handler" to equal "file" (https:/
This is done so that libvirt operates like it is less than version 1.3.0 and does not look for a new daemon that was introduced in 1.3.0. This is what happens when 1.3.0 is installed without this option: https:/
The option was added to Kolla here: https:/
However from what I'm seeing this is not being set because the template for /etc/libvirt/
" error mentioned in the redhat bug linked above.
Assuming this assessment is correct the nova_libvirt Dockerfile template should be changed to set "stdio_handler" as desired if the option is present BUT also add the desired line if there is not existing value.
some info on the environment in which I'm seeing this:
branch in use: stable/mitaka
distro: Ubuntu 15.10
install type: source
multinode
KVM enabled / non qemu
Changed in kolla: | |
importance: | Undecided → Critical |
summary: |
- nova_libvirt's Dockerfile.j2 has a sed one-liner which does not work - because the target file is empty + recent change to qemu.conf breaks some deployments |
Changed in kolla: | |
status: | New → Confirmed |
assignee: | nobody → Steven Dake (sdake) |
milestone: | none → newton-1 |
I forgot to mention that if I use ansible to add this option on my target nodes and restart nova_libvirt container I am then able to start VMs without error.
ansible -i /usr/local/ share/kolla/ ansible/ inventory/ multinode -m shell -a 'echo "stdio_handler = \"file\"" >> /etc/kolla/ nova-libvirt/ qemu.conf' all