#atlantis (2020-09)

https://www.runatlantis.io/docs/locking.html#unlocking

if I unlock it will cancel the current running apply?

In my case, yes, github stop actions for this job

I see, I will give it a go then thanks

I think you can also comment atlantis unlock

it would be really nice to get more logs from atlantis

True, even with debug there is not a lot to read.

thanks for your help - maybe there are good alternatives with this functionality

No there is currently no way to stream the output from the plan/apply to logs

Trick is to make sure what Atlantis is running, is exactly what you can run locally.

right, makes sense - we were having problems because during the apply it was trying to wget the EKS private API

which the atlantis machine did not have permission to do

Do you know if Atlantis is still being actively maintained? (looks to be from the github repo)

I think Luke has mostly stopped with new features but maybe helping with PRs (since joining Hashicorp himself) but community is mostly doing bug fixes / any new features

There was talk about an API endpoint to run plan/apply that sounded like a nice addition, not sure how far that got.

want https://github.com/runatlantis/atlantis/pull/997

right, I guess he now works for his own competition

perhaps he will hand over ownership if someone is willing ..

Well, they hired him to work on a similar product, so pretty sure they aren’t too keen on him working on an open source alternative

https://github.com/runatlantis/atlantis/pull/997#issuecomment-675767198 28 days ago, maybe soon?

ah cool

Padarn Wilson ?

You look like another Padarn I know from a another Slack

I am another Padarn you know from slack haha

small world

?

+

Fun times.

There are lots of workable approaches, but I think using a single atlantis service that can assume roles is the more standard approach

One service is not acceptable , legal stuff.

That makes things a bit harder, but I think your approach makes sense with that requirement

Right now, two separated services deployed. Prod service complains

parsing atlantis.yaml: workflow "stage" is not defined anywhere

so I think I need to add dummy workflow for stage on production service.

I run multiple atlantis in different accounts

you will get that error no matter what

I use this fork

https://github.com/sonatype/atlantis

look at the code changes, there is one specifically so that the atlantis server only parse atlantis-prod.yaml, atlantis-stage.yaml and does not try to run plan

that way you just get the error you mention only

I fixe the issue by adding to production atlantis service dummy stage workflow with true on plan and apply like this

  repo-config-json = {
    repos = [
      {
        id                     = "/.*/",
        allowed_overrides      = ["apply_requirements", "workflow"],
        allow_custom_workflows = true
        apply_requirements     = ["mergeable"]
      }
    ]
    workflows = {
      prod = {
        plan = {
          steps = [
            {
              run = "terragrunt plan -no-color -out $PLANFILE 2> $PLANFILE.stderr || cat $PLANFILE.stderr"
            }
          ]
        },
        apply = {
          steps = [
            {
              run = "terragrunt apply"
            }
          ]
        }
      },
      stage = {
        plan = {
          steps = [
            {
              run = "true"
            }
          ]
        },
        apply = {
          steps = [
            {
              run = "true"
            }
          ]
        }
      }
    }
  }

anyone know why they don’t allow to set yaml file per instance?

nice hack!!! that is cool, I did not bother to fix the errors since they added an option to delete old comments so the error shows for very little time and then it goes away

I have been following that request a lot and the answers about multiserver have been very simplistic in my opinion and somehow the devs think is more complicated that what you did or what I did

things like “smarter way” have been mention in the PRs with no explanation of the idea behind

at the same time the Dev that maintains Atlantis works in hashicorp so there might be conflicts of interests

it is funny that this kind of features are not added and there is like ~350 forks of atlantis

it’s not working , but I’ve decided to apply patch to original source it’s just

diff --git a/server/events/yaml/parser_validator.go b/server/events/yaml/parser_validator.go
index 0e2bd84a..048c7efd 100644
--- a/server/events/yaml/parser_validator.go
+++ b/server/events/yaml/parser_validator.go
@@ -17,7 +17,16 @@ import (
 )

 // AtlantisYAMLFilename is the name of the config file for each repo.
-const AtlantisYAMLFilename = "atlantis.yaml"
+var AtlantisYAMLFilename string
+
+// Simplest hack to allow overriding "atlantis.yaml" to another name
+func init() {
+       AtlantisYAMLFilename = os.Getenv("ATLANTIS_YAML_FILENAME")
+       if AtlantisYAMLFilename == "" {
+               AtlantisYAMLFilename = "atlantis.yaml"
+       }
+}
+

Finally without patch, one instance of Atlantis for 3 account

  repo-config-json = {
    repos = [
      {
        id                     = "/.*/",
        allowed_overrides      = ["apply_requirements", "workflow"],
        allow_custom_workflows = true
        apply_requirements     = ["mergeable"]
      }
    ]
    workflows = {
      cicd = {
        plan = {
          steps = [
            {
              run = "terragrunt plan -no-color -out $PLANFILE 2> $PLANFILE.stderr || cat $PLANFILE.stderr"
            }
          ]
        },
        apply = {
          steps = [
            {
              run = "terragrunt apply"
            }
          ]
        }
      },
      prod = {
        plan = {
          steps = [
            {
              "env" : {
                "name" : "TERRAGRUNT_IAM_ROLE",
                "value" : "${local.production_role}"
              }
            },
            {
              run = "terragrunt plan -no-color -out $PLANFILE 2> $PLANFILE.stderr || cat $PLANFILE.stderr"
            }
          ]
        },
        apply = {
          steps = [
            {
              "env" : {
                "name" : "TERRAGRUNT_IAM_ROLE",
                "value" : "${local.production_role}"
              }
            },
            {
              run = "terragrunt apply"
            }
          ]
        }
      },
      stage = {
        plan = {
          steps = [
            {
              "env" : {
                "name" : "TERRAGRUNT_IAM_ROLE",
                "value" : "${local.staging_role}"
              }
            },
            {
              run = "terragrunt plan -no-color -out $PLANFILE 2> $PLANFILE.stderr || cat $PLANFILE.stderr"
            }
          ]
        },
        apply = {
          steps = [
            {
              "env" : {
                "name" : "TERRAGRUNT_IAM_ROLE",
                "value" : "${local.staging_role}"
              }
            },
            {
              run = "terragrunt apply"
            }
          ]
        }
      }
    }
  }
}

why the patch did not work?

Atlantis confused developers, one instance commented Ran plan for 0 projects and the other commented with proper plan

yes that is a side effect

so now you get one plan that works and one that does not in your comments?

now I have one instance of Atlantis which use switch role to access stage/prod env

we have separated CI/CD account in which Atlantis is deployed, one atlantis yaml with 3 workflows, basically Atlantis operates on CICD/stage/prod account

ahhhh i c ok

dang! still no update on this, huh? 34 too!

yep, still waiting