Vagrant Up Works But Deploying to Droplet Gets Errors

Trying to deploy to a staging environment with ./bin/ staging ‘’ fails trying to get private packages from packagist (authentication). (vagrant up seems to work.) These same playbooks and associated setup work for my colleagues. If I check auth.json on the droplet (remote VM) the file has no credentials. If I pause the playbook and add a packagist token on the remote server I can avoid the error but the resulting droplet still doesn’t work normally.

I have a suspicion that more than just the token is failing to make it in to the environment when I deploy to staging, but the auth errors trying to get to the private repo are where the playbook fails explicitly.

I am looking for ideas. I clone the same files and appear to have the same tools/software installed as my coworkers, but my deploys to staging fail. Note that after the deploy fails with the auth errors with repo.packagist, the droplet has no current directory like /srv/www/name/current where I would normally see elements of the install.

Any suggests or comments are welcome and appreciated!

:thinking: that’s very odd. If your colleagues can deploy without issue I can’t think of why you couldn’t. Assuming, as you said, that everyone is using the same playbooks from source control.

droplet has no current directory like /srv/www/name/current

This should only happen if there’s never been a successful deploy (or someone manually deletes it). If there’s a current release, failures would just prevent that symlink from being updated to the new release, and not delete it.

Hey Scott,

Thanks a bunch for the reply. I am trying to deploy into a fresh droplet to support user testing. I have been destroying and trying again, so they haven’t had a successfully deploy yet (this is not true of my colleagues systems or production). If I pause the playbook in build-after.yml after task “Setup composer authentications” and go to the droplet and manually add the packagist token to auth.json before I allow it to continue, the errors in “Install Composer dependencies” don’t happen and current is created, but the site in the staging droplet still doesn’t come up.


You can get Trellis to generate an auth.json for you on deploy: Composer failing due to incorrect GitHub fingerprint - #7 by alwaysblank

You mention this is a “fresh droplet”: have you provisioned it? Trellis: Remote Server Setup | Roots Documentation

Hi Ben,

Thanks for the help! Your link told me that others had occasionally seen similar issues it really helped knowing I wasn’t on a wild goose chase.

Our vault has packagist and github tokens which I can see with ansible-vault view, but my deploys don’t seem to pick them up. My remote auth.json is empty json. I had previously tried adding my packagist token manually to the remote auth.json and that removed errors, but still didn’t result in a working site. After reading your suggestions I transferred my entire local auth.json which included a github token as well as a packagist token and the remote site came up normally. Finally!

I really appreciate the help. I don’t know why the tokens aren’t pulled from the vault (they do get pulled successfully for my coworkers), but transferring the auth.json is an easy extra step.

Do you have the appropriate user keys? trellis/users.yml at master · roots/trellis · GitHub

This topic was automatically closed after 42 days. New replies are no longer allowed.