Skip to content

Fix ssh key for bootstrapping - #133

Open
sabaini wants to merge 1 commit into
canonical:mainfrom
sabaini:bug/2054328
Open

sabaini wants to merge 1 commit into
canonical:mainfrom
sabaini:bug/2054328

Conversation

@sabaini

@sabaini sabaini commented Feb 19, 2024

Copy link
Copy Markdown

Create a dedicated passwordless ssh key for bootstrapping the local juju controller

Fixes https://bugs.launchpad.net/snap-openstack/+bug/2054328

@sabaini
sabaini force-pushed the bug/2054328 branch 2 times, most recently from 362566c to e6feb8a Compare February 20, 2024 10:22
@hmlanigan

Copy link
Copy Markdown

@sabaini, thank you for the pull request. If this change is still valid, please resolve the conflicts and we'll get this reviewed/landed.

@sabaini
sabaini force-pushed the bug/2054328 branch 2 times, most recently from 33d2493 to f7d0b31 Compare July 7, 2026 15:31
@sabaini
sabaini force-pushed the bug/2054328 branch 3 times, most recently from 88ed2e6 to 633d43c Compare September 1, 2026 14:35
Create a dedicated passwordless ssh key for bootstrapping the local juju
controller. The default key (id_ed25519) cannot be relied upon: it may
already exist and carry a passphrase.

Fixes https://bugs.launchpad.net/snap-openstack/+bug/2054328

Assisted-by: pi:openai/gpt-5.6-sol

Signed-off-by: Peter Sabaini <peter.sabaini@canonical.com>
@gboutry

gboutry commented Sep 4, 2026

Copy link
Copy Markdown
Collaborator

I'm on the fence for this change because I believe it's not a great idea to actually deploy Sunbeam on your work laptop directly. This provides a bit of friction, but it allows the end user to double check what they are doing. It's not a great feature, and we might want to address this another way.

Juju and Sunbeam will take full root ownership of the machine.

@hmlanigan

Copy link
Copy Markdown

@gboutry there may be reasons a user puts an ssh key on the machine being used for sunbeam commands other than it's a work laptop.

Since openstack is a strictly confined snap. Why are we putting data in a user's .ssh directory to begin with? Are we leaving pieces behind when snap remove --purge is used?

Juju and Sunbeam will take full root ownership of the machine.

How does this work in the case of sunbeam bootstrap with an existing juju controller functionality, or the case of a MAAS config? No reason to take over an entire machine in that case.

The prepare script is also too permissive if so. It needs to warn users when unexpected things exist.

We cannot solve the entire problem in this one PR. I'd say that prepare sunbeam is doing the wrong thing per ssh files.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants