The post explains how to fix a potential network problem when using a CI template. It is based on a true story after creating an Ubuntu 24.04 template and using it.

Use case

ci.inria.fr allows users to create their own templates. The process is quite simple:

• create a slave as usual
• configure it as you wish (installing your preferred packages, tools, environment, conda, etc.

• connect to Cloudstack

  • stop the slave
  • create a template from this slave (the process and instructions on where to click can be found in the Cloudstack documentation)

This is exactly what I did for a Ubuntu 24.04 slave.

Then, I created a new slave using this homemade template…

And… patatras !

The problem

The new slave started, but it was not reachable online (SSH connections failed).

The fix

• connect to the Cloudstack console for this slave (before this, I had to force reboot the slave several times, for reasons still unknown).
• log in using the ci account.
• increase your privileges to root (sudo -i).
• verify that the network interface is broken (e.g., by using ifconfig).
• get the MAC address of the network interface (ifconfig will help you again; for Ubuntu 24.04, the interface is ens3).
• modify the file /etc/netplan/50-cloud-init.yaml: replace the MAC address in the file with the actual MAC address of the current slave.
• reboot the slave.

Et voilà!

More cleaning

The new slave inherits all the characteristics of the original one. Some additional fixes can/should be performed:

As root:

• change the hostname of the new slave (echo new_slave_hostname > /etc/hostname).
• regenerate the SSH server keys (dpkg-reconfigure openssh-server).

As the ci user:

• update the GitLab/GitHub runner credentials if they are part of the template (remove the old ones, create new ones).

PS: This list might grow if new issues are discovered with the new slave…