Skip to content
This repository has been archived by the owner on Feb 5, 2020. It is now read-only.

Installation hangs on null_resource.kubeconfig #537

Closed
jpward1981 opened this issue May 4, 2017 · 9 comments
Closed

Installation hangs on null_resource.kubeconfig #537

jpward1981 opened this issue May 4, 2017 · 9 comments

Comments

@jpward1981
Copy link

When attempting to run through the installation process on some bare metal hosts. I run into the issue where the installation process would hang on the null_resource.kubeconfig.* step and it would timeout after an hour. I know that during this process, this is when the installer is waiting for the host installation to complete. Once complete the kubeconfig is copied to the hosts via ssh.

It appears that the installer can't connect to the hosts via ssh. On the Container Linux host I can see auth failures repeated until the process times out. In my configuration file, I have provided a public key and i can ssh to the hosts after the process times out.

@dghubble
Copy link
Member

dghubble commented May 4, 2017

Is this with the graphical installer or running terraform apply manually? What's your local SSH key setup?

In the graphical installer, terraform runs in it's own env and we fixed an issue to pass SSH_AUTH_SOCK (Yubikeys, etc.). Otherwise, I believe terraform looks for common key locations. What version of this repo did you try?

@jpward1981
Copy link
Author

I was running terraform manually. I am able to ssh directly to the servers using the public key/private key properly. I was using the latest master.

@dghubble
Copy link
Member

dghubble commented May 9, 2017

@coreypobrien
Copy link
Contributor

It appears Terraform uses ssh-agent, so if it isn't configured or available, the only options attempted are the password or private_key provided in the connection block. The key lookups that the openssh client does (e.g. ~/.ssh/id_rsa) don't happen.

PR #625 adds an optional variable to specify the private_key in cases where ssh-agent isn't available.

@steigr
Copy link

steigr commented May 16, 2017

I had the same problem (using GUI installer). I got rid of it by removing all unneeded identities from the ssh-agent instance. (ssh-add -D; ssh-add ~/.ssh/id_ed25519)

@zendeavor
Copy link

Confirming this issue affected me as well. I wasn't running ssh-agent at all, so once I started it up on the system from where I'm running the tectonic installer and pointed it toward my key, kubeconfig was able to do its work.

@robszumski
Copy link
Member

Added a tracking issue for documentation improvements here: #1078

We can keep this open until that lands.

@alexandrufulop
Copy link

more clear info here: #1078

@dghubble dghubble removed their assignment Oct 13, 2017
@dghubble
Copy link
Member

Closing as #1078 documents this as a requirement.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Projects
None yet
Development

No branches or pull requests

8 participants