text file busy\

[13567436138]

{"level":"error","ts":"2025-12-06T05:16:22.272Z","caller":"events/instrumented_project_command_runner.go:78","msg":"Error running plan operation: running 'sh -c' '/atlantis-data/bin/terraform1.14.1 init -input=false -upgrade' in '/atlantis-data/repos/atlantis/k8s-nginx/9/default': exit status 1\nInitializing the backend...\nInitializing provider plugins...\n- Finding hashicorp/kubernetes versions matching \"~> 2.38.0\"...\n- Installing hashicorp/kubernetes v2.38.0...\n╷\n│ Error: Failed to install provider\n│ \n│ Error while installing hashicorp/kubernetes v2.38.0: open\n│ /atlantis-data/plugin-cache/registry.terraform.io/hashicorp/kubernetes/2.38.0/linux_amd64/terraform-provider-kubernetes_v2.38.0_x5:\n│ text file busy\n╵\n","json":{"repo":"atlantis/k8s-nginx","pull":"9"},"stacktrace":"github.com/runatlantis/atlantis/server/events.RunAndEmitStats\n\tgithub.com/runatlantis/atlantis/server/events/instrumented_project_command_runner.go:78\ngithub.com/runatlantis/atlantis/server/events.(*InstrumentedProjectCommandRunner).Plan\n\tgithub.com/runatlantis/atlantis/server/events/instrumented_project_command_runner.go:38\ngithub.com/runatlantis/atlantis/server/events.runProjectCmds\n\tgithub.com/runatlantis/atlantis/server/events/project_command_pool_executor.go:48\ngithub.com/runatlantis/atlantis/server/events.(*PlanCommandRunner).run\n\tgithub.com/runatlantis/atlantis/server/events/plan_command_runner.go:272\ngithub.com/runatlantis/atlantis/server/events.(*PlanCommandRunner).Run\n\tgithub.com/runatlantis/atlantis/server/events/plan_command_runner.go:321\ngithub.com/runatlantis/atlantis/server/events.(*DefaultCommandRunner).RunCommentCommand\n\tgithub.com/runatlantis/atlantis/server/events/command_runner.go:427"}

[dosubot[bot]]

The "text file busy" error during Terraform provider installation in Atlantis is a known concurrency issue. It happens when multiple parallel Terraform runs (like with parallel_plan or parallel_apply enabled) try to read/write the same provider binary in the shared plugin cache at the same time. The plugin cache directory (TF_PLUGIN_CACHE_DIR) is not concurrency safe, so this race condition can cause errors like the one you saw reference.

To work around this, you have a few options:

• Disable the plugin cache by starting Atlantis with --use-tf-plugin-cache=false. This avoids the race but will increase provider download times reference.
• Serialize the terraform init step using a lock (e.g., with flock) in your Atlantis workflow, so only one init runs at a time:

  workflows:
    default:
      plan:
        steps:
          - run: flock /tmp/terraform_init.lock bash -c "rm -rf .terraform/providers && env -u TF_PLUGIN_CACHE_DIR TF_CLI_CONFIG_FILE=/dev/null terraform init -upgrade"
          - plan

• Use unique plugin cache directories per project, or advanced solutions like OverlayFS (requires privileged containers) reference.
• Setting TF_PLUGIN_CACHE_MAY_BREAK_DEPENDENCY_LOCK_FILE=true can help in some cases, but doesn't fully resolve the race.

There’s currently no built-in Atlantis feature to serialize only the init step while keeping plan/apply parallel reference. Disabling the plugin cache is the simplest workaround for large parallel deployments, but it does impact performance.

Let me know if you want more details on any of these approaches.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}