Can deploy to production, but not staging

[encryptdesigns]

FWIW I followed the same steps as Josh did in his first post here:

Can't deploy: TASK [deploy : Clone project files] - git repo cannot be accessed

First let me say that yes, I’ve searched this discourse and Google and I’m unable to find a solution to this problem. I’ve regenerated keys, I’ve created new github repos, I’ve spooled up new DO droplets and no matter what I attempt to try to do to resolve this, when I get to: TASK [deploy : Clone project files] I receive the following error: Cloning into '/srv/www/-redacted-.com/shared/source'... git@github.com: Permission denied (publickey). fatal: Could not read from remote repository. Pl…

[encryptdesigns]

Should running evalssh-agent -s`` return a new agent pid 26167? Each time I run it I get a new pid. I’m not sure if that’s expected behavior.

[encryptdesigns]

Also when I $git log -- trellis I don’t show anything being changed there since before deployments to staging stopped working. I realize that not all files were checked into git but trying to determine what had changed and if that change broke staging deployments.

[codepuncher]

I’ve seen people with the same issue before and it was because they were deploying a non-existent branch. Have you checked which branch you’re deploying and that it exists on your remote repo?

[encryptdesigns]

Thanks for the reply. The branch is staging and hasn’t changed. I think it’s likely something with my local. I checked SSH Forwarding, SSH Agent, modified my ssh config, and more.

What else is left? ¯_(ツ)_/¯

[alwaysblank]

Can you SSH in to the server and pull down the repo manually? Not a solution, but it could help diagnose where the problem is occuring.

[ng3]

Has the staging branch been pushed up to GitHub?

edit:

1. Sorry, just saw that you said production uses the same branch - so sanity check here: Are you deploying your staging branch to production with no problems?
2. Have you ever been able to deploy to this staging environment? Did provisioning run smoothly?

[encryptdesigns]

1. I meant to say that we have been using the staging for our staging server. While master is used for production. We were able to deploy just fine the other day. I think something on the server got messed up somehow, but I’m not sure what else to try.
2. Yes and yes.

Thanks for all of your help!

[encryptdesigns]

Doesn’t look like that’s an option here since git isn’t deployed to the server.

[alwaysblank]

Trellis deploys by (in part) git cloneing your repository on your remote server (see: https://roots.io/trellis/docs/remote-server-setup/#deploy). If you don’t have git on your remote server, it would be impossible for it to do that.

[encryptdesigns]

Thanks for the response. I should have said that git itself isn’t installed, but I do understand that trellis utilizes that to deploy the updates. The same can be said of production too, where deployments work just fine.

Is there anything that I can check that may have been foobarred on the server?

[encryptdesigns]

I can successfully provision with $ ansible-playbook server.yml -e "site=SITENAME.com env=staging" but not deploy $ ansible-playbook deploy.yml -e "site=SITENAME.com env=staging"

Any other ideas besides provisioning another fresh server?

[ng3]

Ok, have you tried deploying to staging with your master branch?

Can you show us the contents of your trellis/group_vars/staging/wordpress_sites.yml please?

[TangRufus]

git should be installed during provision - https://github.com/roots/trellis/blob/49cf5de36c7bfb0661a922288da898228cc526e9/roles/common/defaults/main.yml#L27

---

Change https://github.com/roots/trellis/blob/49cf5de36c7bfb0661a922288da898228cc526e9/roles/deploy/tasks/update.yml:

  - name: Clone project files
    git:
      repo: "{{ project_git_repo }}"
      dest: "{{ project_source_path }}"
      version: "{{ project_version }}"
      accept_hostkey: "{{ project.repo_accept_hostkey | default(repo_accept_hostkey | default(true)) }}"
      force: yes
-   ignore_errors: true
-   no_log: true
+   ignore_errors: false
+   no_log: false
    register: git_clone

Then, re-deploy with $ ansible-playbook deploy.yml -e "site=MYSITE.com env=staging" -vvv , you will get a better error.

[encryptdesigns]

@ng3

trellis/group_vars/staging/wordpress_sites.yml

Here it is:

wordpress_sites:
  SITENAME.com:
    site_hosts:
      - canonical: staging.SITENAME.com
    local_path: ../site # path targeting local Bedrock site directory (relative to Ansible root)
    repo: git@github.com:SITENAME/SITENAME.com-wordpress.git # replace with your Git repo URL
    repo_subtree_path: site # relative path to your Bedrock/WP directory in your repo
    branch: staging
    multisite:
      enabled: false
    ssl:
      enabled: true
      provider: letsencrypt
      hsts_include_subdomains: false
    cache:
      enabled: false
    htpasswd:
     - name: SITENAME
       password: staging

[encryptdesigns]

@TangRufus

• ignore_errors: true - no_log: true + ignore_errors: false + no_log: false

After updating that file I get this:

TASK [deploy : Clone project files] **********************************************************************************************************************************************************************************************************************
System info:
Ansible 2.5.3; Darwin
Trellis version (per changelog): "Update xdebug tunnel configuration"
---------------------------------------------------
Local modifications exist in repository (force=no).
fatal: [staging.SITENAME.com]: FAILED! => {"before": "b0af714b56ce0f0b39ef303b653d18dbc04ad2c5", "changed": false}

Does that help determine the issue at all?

[encryptdesigns]

Updating force, per this link, looks like it corrected my issues:

- name: Clone project files
  git:
    repo: "{{ project_git_repo }}"
    dest: "{{ project_source_path }}"
    version: "{{ project_version }}"
    accept_hostkey: "{{ project.repo_accept_hostkey | default(repo_accept_hostkey | default(true)) }}"
    force: yes
  ignore_errors: false
  no_log: false
  register: git_clone

I was able to deploy to staging. I have since removed the force option, and can still deploy!

Thank you all for the help in pointing me in the right direction!

[TangRufus]

For those who coming here from google, the reason is that /srv/www/XXXX/shared/source being modified manually (which you should never do!)

To make deploy works again:

Option 1: Set force: true to discard local changes caught by git
Option 2: Delete /srv/www/XXXX/shared/source
Option 3 (best): Update Trellis, specifically this pull request - https://github.com/roots/trellis/pull/999

No matter which option you go for, don’t forget reverting to no_log: true as logging sometimes prints out git credentials.

Rule of thumb: If you are changing remote server source code manually, you probably doing it wrong.

[encryptdesigns]

Completely agree. I think panic mode crept in when we had an issue with the site, then at the same time deployments w/staging occurred. Sometimes it’s better to take a step back. #LessonLearned.

Thanks again all!