isaaguilar
commented
2 years ago I was able to replicate the issue on the ARM64v8 architecture, but not on the AMD64. I'll have to dig to see why the different builds have different results for ssh keys.
I ran a test as similar as possible to yours. This is the config that works for amd and not for arm:
```yaml apiVersion: tf.isaaguilar.com/v1alpha2 kind: Terraform metadata: name: example namespace: default spec: terraformVersion: 1.0.0 # Pull this module to execute terraformModule: source: "git@github.com:isaaguilar/simple-aws-tf-modules.git//private-github-module" # Use kubernetes as a backend which is available for terraform >= v0.13 backend: |- terraform { backend "s3" { region = "us-east-1" bucket = "my-terraform-state-bucket" key = "terraform-operator/my/awesome/example.tfstate" } } ignoreDelete: true # Create a tfvar env for the terraform to use taskOptions: - for: - '*' # The following config affects all task pods env: - name: TF_VAR_vpc_name value: blah - name: TF_VAR_client_id value: acme - name: TF_VAR_environment value: test envFrom: - secretRef: name: aws-session-credentials # temp creds for my bucket scmAuthMethods: - git: ssh: sshKeySecretRef: key: key name: gitsshkey namespace: default host: github.com keepCompletedPods: true keepLatestPodsOnly: true serviceAccount: tf-operator-service-account ``` The main module is simple: ```hcl # https://github.com/isaaguilar/simple-aws-tf-modules/blob/master/private-github-module/main.tf output "static" { value = "static" } module "private" { // This source is private source = "git@github.com:isaaguilar/terraform-do-something-awesome.git?ref=main" } terraform { required_version = "> 0.12" } ``` And indeed my logs match yours: ``` Connecting to raw.githubusercontent.com (185.199.111.133:443) saving to '/home/tfo-runner/generations/1/init.sh' init.sh 100% |********************************| 2762 0:00:00 ETA '/home/tfo-runner/generations/1/init.sh' saved Initializing modules... Downloading git@github.com:isaaguilar/terraform-do-something-awesome.git?ref=main for private... ╷ │ Error: Failed to download module │ │ Could not download module "private" (main.tf:7) source code from │ "git@github.com:isaaguilar/terraform-do-something-awesome.git?ref=main": │ error downloading │ 'ssh://git@github.com/isaaguilar/terraform-do-something-awesome.git?ref=main': │ /usr/bin/git exited with 128: Cloning into '.terraform/modules/private'... │ error: cannot run ssh: No such file or directory │ fatal: unable to fork │ ╵ ```Any help determining why the arm build isn't working is greatly appreciated.
odise
I noticed some weird behaviour when TF code to be deployed is depending on private GIT repositories. The
terraform initstep fails with the following error:Interestingly the
setupstep runs successful although it depends on an internal GIT repository. The error above is a result ofterraform inittrying to download further modules from the same GIT repo.Here is my Terraform manifest:
I defines a
preinitstep to investigate the/home/tfo-runner/.sshdirectory and found this:All attempts to change the permissions within
preinitfailed ininitwith same result.I'm using the Helm chart v0.2.15 from https://galleybytes.github.io/helm-charts which installs
terraform-operator:v0.9.0-pre3.