Cannot set Server Configuration values that are only settable through config file

[dennislapchenko]

Slack token, for example, is currently only configured through atlantis configuration file. Any idea how to implement it with current setup?

Smoothest option is to add another cloudinit.write_files entry, right?

[bschaatsbergen]

Hi @dennislapchenko,

According to https://www.runatlantis.io/docs/using-slack-hooks.html - it should be possible to set the Slack token through setting the environment variable: ATLANTIS_SLACK_TOKEN

And I believe that you can set the Slack configuration through the repo/server yaml config. This module supports passing the server config, see /examples. And for the repo config, you can set the Slack config as part of that atlantis.yaml (in your repository).

Hopefully this helps you, if not let me know!

[dennislapchenko]

@bschaatsbergen thanks for a swift reply. At first I also thought that its the server yaml config that it should go to. But after trying it out - came to understand its the actual atlantis config file that this has to be in. (path to which is through ATLANTIS_CONFIG). webhooks is not declared in server atlantis yaml schema (nor repo atlantis yaml, but its not even important)
This issue comment proves it: runatlantis/atlantis#2143 (comment)

[dennislapchenko]

Before I attempted a fork or a PR into cloudinit, I tried to use startup_script like:

  env_vars = {
    ATLANTIS_CONFIG = "/home/atlantis/config.yaml"
  }

  startup_script = <<EOF
#!/usr/bin/env bash

cat <<YAML > $ATLANTIS_CONFIG
webhooks:
  - event: apply
    workspace-regex: .*
    kind: slack
    channel: terraform-notifications
YAML
EOF

But i do get:

Error: invalid config: reading /home/atlantis/config.yaml: open /home/atlantis/config.yaml: no such file or directory

It seems the container is starting faster than the startup script

[bschaatsbergen]

Hmm thanks for elaborating. I'll take a stab this 👍🏼 @dennislapchenko

Note: the startup-script is ran on the VM, not within the container itself, the file should be copied into the disk that the container has mounted.

[dennislapchenko]

@bschaatsbergen perhaps adding a cloudinit part {} for config as file is a good option? (Or could just wait until Atlantis team adds webhooks to ENV configurables oto)

Re startup-script - as seen in the example above, i was writing the file to /home/atlantis/config.yaml, and the mounted volume is /home/atlantis so it should've worked right? Perhaps container is launching earlier after all? (interesting that there is not restart after failed startup?)

[bschaatsbergen]

Hi @dennislapchenko, I'll get back on this issue soon. Have you meanwhile find an alternative way to fix it?

[bschaatsbergen]

I plan on checking this out soon @dennislapchenko, been a busy few weeks. Have you tackled this already?

[dennislapchenko]

Hi @bschaatsbergen, unfortunately I havent. Have switched projects and here atlantis is in k8s 😄