OKD cluster creation #26
Conversation
|
Currently, this setup works if you relies on the branch cocl-kcli since it includes the fixes for the TPM and to use a custom url for the coreos image. git clone https://github.com/alicefr/kcli.git -b cocl-kcli
cd kcli
python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -e . |
|
The part that is still missing is how we can customize the ignition files generated by cli for adding the clevis pin |
alicefr
force-pushed
the
okd
branch
2 times, most recently
from
520b4a2 to
107faac
Compare
Integrate the build of the custom scos image. Signed-off-by: Alice Frosi <[email protected]>
|
This looks super interesting. I'm surprised we can get a cluster faster using UPI & libvirt but that is great. If we can get an OKD/OCP UPI cluster up in 5 minutes then that makes it super interesting. |
|
@travier it takes 5 mins if you use the import script from a tarball, not if you create it from scratch. Bootstrapping takes around 30-40 mins, but the export script, it basically dumps the control plane VM, and with the import script you can recreate it |
|
It will be nice to release the tarball somewhere. Right now, it is around 6GB. Unfortunately, the scos image for the control plane is quite large (25-30GB) |
HOWTO creates a local OKD cluster with kcli. Signed-off-by: Alice Frosi <[email protected]>
|
Depend on the scos PR: #27 |
| exit 1 | ||
| fi | ||
|
|
||
| sudo virt-sparsify --in-place "$HOME/VirtualMachines/${VM}_0.img" |
There was a problem hiding this comment.
That path can vary based on system configuration (I don't know what unfortunately), for me it was in /var/lib/libvirt/images.
Just for context (i.e. not criticism) the libvirt interactions also required graphical sudo authentication on my machine even though I have the groups set up to interact with libvirt and create domains as normal user? So it might be something in the kcli config that should be flexibilized or documented here.
There was a problem hiding this comment.
There's some absolute path hardcoding in the result that I'm not sure can be overcome, but that at least I wanted to mention, e.g. when trying yours:
Cannot access backing file
'/home/afrosi/VirtualMachines/scos-9.0.20250510-0-openstack.x86_64.qcow2'
of storage file '/home/jnaucke/VirtualMachines/cocl-ctlplane-0_0.img'
| kcli start plan $CLUSTER_NAME | ||
| echo "Import completed successfully!" | ||
|
|
||
| KUBECONFIG=$HOME/.kcli/clusters/$CLUSTER_NAME/auth/kubeconfig |
There was a problem hiding this comment.
You'll need to export this, otherwise the command below won't work (unless $KUBECONFIG was already exported) and may quietly use some other config
using machine config. Depends on trusted-execution-clusters#26. Towards resolution of github.com/trusted-execution-clusters/operator/issues/13. Signed-off-by: Jakob Naucke <[email protected]>
using machine config. Depends on trusted-execution-clusters#26. Towards resolution of github.com/trusted-execution-clusters/operator/issues/13. Signed-off-by: Jakob Naucke <[email protected]>
There was a problem hiding this comment.
I'd also suggest a bigger disk_size, I ran into disk pressure with the default 30G really fast
3 participants
HOWTO creates a local OKD cluster with kcli.