Wrong home directory for the prometheus user

[ptselios-inmarsat]

The home directory of the prometheus user is /tmp.
Since this is a system user the home directory should either be the directory of the application or the /dev/null.

[paulfantom]

Previously it wasn't set to anything, but after #153 we changed this behavior.

@michalklempa Would setting /dev/null as prometheus user home directory will work on GCE?

[ptselios-inmarsat]

Meaning, if I already have prometheus up and running and run the playbook again, it will fail since the ansible's user module will try to chmod the user's home directory and it will fail since the home directory will be /home/prometheus, it tries to change it to /tmp and prometheus is running...

[paulfantom]

Oh, so this is a problem with an upgrade path and not with result itself, got it. I will add a warning in readme and release notes about this problem with a simplest possible solution:

switch prometheus off before an upgrade

[ptselios-inmarsat]

Hmmm
It's not that simple, unfortunately. I removed my AWS instances and recreated them from scratch.
It failed.
Also, I just run the playbook in order to make sure that the systems are in the desired state. It failed again.
My solution was manually change the entry in the /etc/passwd

[paulfantom]

That is strange. We are running automated daily deployments of our roles against one server (demo.cloudalchemy.org) and it didn't show any signs of problems. As a test I just run this role against vanilla fedora 29 server in digital ocean and it also went ok.

Which ansible version you are using?

[ptselios-inmarsat]

My setup is CentOS 7 in AWS deployed through Ansible Tower 3.3.1

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.