Deploy: OSError: [Errno 13] Permission denied

[strarsis]

So I set up a new staging Ubuntu Server 18.04.2 LTS system on Hyper-V and
successfully provisioned it as trellis system with ansible.

However, deploying a site to it results in an error message very early during deploy:

MODULE FAILURE
Traceback (most recent call last):
File "/tmp/ansible_oJ2ASV/ansible_module_deploy_helper.py", line 529, in
<module>
main()
File "/tmp/ansible_oJ2ASV/ansible_module_deploy_helper.py", line 491, in
main
changes += deploy_helper.create_path(facts['releases_path'])
File "/tmp/ansible_oJ2ASV/ansible_module_deploy_helper.py", line 352, in
create_path
os.makedirs(path)
File "/usr/lib/python2.7/os.py", line 157, in makedirs
mkdir(name, mode)
OSError: [Errno 13] Permission denied: '/srv/www/example.com/releases'
An exception occurred during task execution. To see the full traceback, use -vvv. The error was: OSError: [Errno 13] Permission denied: '/srv/www/example.com/releases'
fatal: [staging-1]: FAILED! => {"changed": false, "module_stdout": "", "rc": 1}

I found similar issues on the internet, so I list them here:

Rmtree failed: [Errno 13] Permission denied at Finalize the deploy

I’ve run into this a few times before - if a prior deploy failed, Ansible will be unable to remove it (the release folder) during a subsequent deploy. Looking at the path referenced, I’d guess your deploy on 12/09/2017 @ 22:27:49 didn’t complete, but the folder is still there. SSH into the machine, and run rm -rf /srv/www/mysite.com/releases/20171209222749/ - you should also run rm -rf /srv/www/mysite.com/releases/20171212123959/ since the deploy failed on that too, so eventually you’ll run i…

https://discourse.roots.io/t/trellis-bin-deploy-sh-deploy-error/9213
https://discourse.roots.io/t/trellis-aws-and-provision/5496/6
https://github.com/roots/trellis/issues/908
https://imwz.io/roots-trellis-errors/

So from what I learned from these discussions is that this kind of error can happen:

• When provisioning hadn’t ran yet, either never had, or for the site to be deployed.
  However, the initial provisioning including that site in config, ran successfully.
• Ownership and permissions of /srv and its subfolders are somehow off.
  These look correct for me, also I can manually create the directory /srv/www/example.com/releases as user web, hence ansible should be able, too.
• In some rare cases something went wrong with the provisioned server and
  the only way fixing is to completely recreate it from scratch.
  This I did after the first occurence of this issue and
  I can do this repeatedly with the same error as result.

I can reproduce this issue each time with a new VM from scratch:

• Create new minimal Ubuntu Server 18.04.2 LTS system (with OpenSSH server) on Hyper-V.
  • Disable IPv6 in Ubuntu because it caused many issues, also for other users.
    So we are using only IPv4 here.
• Install ansible tested with trellis,
  Installed the ansible package available on Ubuntu workstation (ansible 2.4.3.0).
• Clone trellis from GitHub: git clone https://github.com/roots/trellis
  • Set up some example site in group_vars/staging/wordpress_sites.yml
• Run (initial) provisioning, it should complete just fine.
• Now try to deploy the example site,
  it should fail very early during deploy with the error shown above.

How can I find out the reason for (and hopefully fix) this strange issue?

[strarsis]

For devs that encounter this issue in the future:

In hosts/staging ansible_user=admin was set for the host,
this caused the issue. This also seems to be a bug because not a single source of truth is used for the username, or at least the username can be partly influenced by ansible_user.