Revert "templates: Disable SSH keys lookup from authorized_keys.d"#2283
Revert "templates: Disable SSH keys lookup from authorized_keys.d"#2283LorbusChris wants to merge 1 commit intoopenshift:masterfrom
Conversation
This reverts commit 5ed6fa3.
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: LorbusChris The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
|
/test okd-e2e-gcp-op |
|
@LorbusChris: The following tests failed, say
Full PR test history. Your PR dashboard. DetailsInstructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here. |
|
Didn't we agree there's an important security issue being fixed by that change? I think we need a plan other than just reverting. I don't fully understand the problem here honestly - actually, how can the MCO affect SSH access to the bootstrap? Isn't all the Ignition for that generated by the installer? Would it somehow address the problem to move writing this file into the MCD-as-daemonset logic instead of Ignition? |
|
@cgwalters the MCO doesn't affect the bootstrap node - it's just that we do not currently have SSH access to any node before MCO starts to manage the key, as Ignition on FCOS writes it to |
|
I wonder, how is ignition configured on RHCOS to write to |
|
I think @cgwalters's point is that Ignition on RHCOS writes to |
oh that's right 🤦♂️, I'll close this PR.
@bgilbert does Ignition have any runtime config options? Could we make this into one? |
We generally avoid them, and there's very little opportunity to invoke one anyway, since the initramfs environment is mostly not customizable by the user. Perhaps we should revisit writing the |
This reverts commit 5ed6fa3.
Not having SSH access to the bootstrap node/before MCO starts successfully on OKD is hindering the ability to debug installation related failures immensely.
We'll need to find another way to solve this properly.
/cc @bgilbert