COI N1KV deployment: Control Kernel Version on compute nodes to a single one.
Today in COI deployment, we noticed that kernel version is different across compute nodes. It picks the latest when the compute node is brought up (as on toady 3.2.51). Instead of this, we need all compute nodes to come up with a single kernel version. Please let us know if this is possible. We will let you know on the kernel version to be locked shortly.
Blueprint information
- Status:
- Complete
- Approver:
- Mark T. Voelker
- Priority:
- Medium
- Drafter:
- Kiran Chunduri
- Direction:
- Approved
- Assignee:
- Mark T. Voelker
- Definition:
- Approved
- Series goal:
- Accepted for grizzly
- Implementation:
-
Implemented
- Milestone target:
-
g.2
- Started by
- Mark T. Voelker
- Completed by
- Mark T. Voelker
Related branches
Related bugs
Sprints
Whiteboard
I would think that the right thing to do would not be to force a particular kernel version to be loaded, but rather to allow the user to specify the kernel package he wants installed (the default being linux-image-
The refined procedure to be added to the late_command sequence looks something like this:
in-target /usr/bin/apt-get install -y $::load_kernel_pkg
export kernel_ver=`echo '$::load_
IFF there is a newer kernel present on the box than the one the user wants:
in-target /bin/sed -i \"/GRUB_DEFAULT=/ s/[0-9]/\"Previous Linux versions>Ubuntu, with Linux \$kernel_ver\"/\" /etc/default/grub ;
Else:
in-target /bin/sed -i \"/GRUB_DEFAULT=/ s/[0-9]/\"Ubuntu, with Linux \$kernel_ver\"/\" /etc/default/grub ;
in-target /usr/sbin/
Pull request issued: https:/
