Terraform Enterprise fails to start when /var or Docker root directory mount use the noexec option – HashiCorp Help Center

Problem

Terraform Enterprise fails to start when either the /var or Docker root directory mount(docker info -f '{{ .DockerRootDir}}') use the noexec option.

Prerequisites

Terraform Enterprise

Cause

The noexec option is set for /var or Docker root directory in /etc/fstab, example:

UUID=920cd144-09bc-487c-ad56-0f09fdd53cb9 /var xfs nodev,noexec,nosuid 0 2
UUID=cd2822cd-6ed5-4bda-9912-eb013f62d763 /var/lib/docker xfs nodev,noexec,nosuid 0 2

If the noexec option has been applied to the mount before Terraform Enterprise is installed the failure happens during the bootstrapping phase:

# tfe-bootstrap
/tfe-bootstrap.sh: line 12: /build-worker-metadata-firewall.sh: Permission denied

# TFE startup failed
[
  {
     "AppID": "0129182acab44ab5534b4b6205257550",
     "Sequence": 722,
     "PatchSequence": 0,
     "State": "stopped",
     "DesiredState": "started",
     "Error": "Container tfe-bootstrap failed: Container 31da3bdbdcbdb79ef7513eda30cdd9f968e46b84e26c6d20be13a39144696ea8 exited with non-zero exit status 126: ",
     "IsCancellable": false,
     "IsTransitioning": false,
     "LastModifiedAt": "2023-07-28T14:17:30.278240957Z"
  }
]

When the noexec option is applied after installing Terraform Enterprise there is a startup failure in multiple containers including tfe-vault, tfe-health-check, tfe-state-parser, tfe-backup-restore and tfe-base-startup. Finally the Terraform Enterprise application startup will timeout leaving it in a failed state.

# tfe-vault
...
+ starting vault
...
Vault is already initialized
+ killing vault with pid 23
==> Vault shutdown triggered
+ vault has exited
+ exiting vault setup with 0
+ Retrieving Vault unseal key
+ Retrieving Vault root token
+ Setting IPC lock...
/usr/bin/vault-start: line 182: /gosu/gosu: Permission denied
+ Starting vault

# tfe-health-check
/usr/bin/setup-ca-certificates.sh: exec: line 41: /gosu/gosu: Permission denied

# tfe-state-parser
/usr/bin/setup-ca-certificates.sh: exec: line 41: /gosu/gosu: Permission denied

# tfe-backup-restore
INFO: Vault token retrieval timeout not yet reached
ERROR: Operation timed out waiting for vault token

# tfe-base-startup
INFO: Vault token retrieval timeout not yet reached
ERROR: Operation timed out waiting for vault token

# TFE startup failed
[
  {
     "AppID": "6e613d8ecae148c3642c6588fe75b597",
     "Sequence": 722,
     "PatchSequence": 0,
     "State": "stopped",
     "DesiredState": "started",
     "Error": "Container tfe-base-startup failed: Container 56bd2bc2dcd8f523e281923354fe41fd913e0c12f9f569fdd82c3a5115eb0d23 exited with non-zero exit status 1: ",
     "IsCancellable": false,
     "IsTransitioning": false,
     "LastModifiedAt": "2023-07-19T15:37:12.954203171Z"
  }
]

Solution #1:

Stop the Terraform Enterprise application, Replicated services, Docker and associated services.

# Stop TFE
replicatedctl app stop
# Verify it has successfully stopped before proceeding
replicatedctl app status

# Stop Replicated services
sudo systemctl stop replicated replicated-operator replicated-ui

# Stop Docker
sudo systemctl stop docker

# Stop Docker socket
sudo systemctl stop docker.socket

# Stop containerd
sudo systemctl stop containerd

Edit /etc/fstab and remove the noexec option from /var or Docker root directory mount, save the file changes and remount the partition.

# Use your favorite editor, example: vim
sudo vim /etc/fstab

# Example entry in fstab for /var/lib/docker
UUID=cd2822cd-6ed5-4bda-9912-eb013f62d763 /var/lib/docker xfs nodev,nosuid 0 2

# Remount the partition
sudo mount -o remount /var/lib/docker

Start Docker, verify Replicated has started successfully and start Terraform Enterprise

# Start Docker
sudo systemctl start docker

# Verify Replicated is ready before proceeding
replicatedctl system status

# Start Terraform Enterprise
replicatedctl app start

Solution #2:

If to comply with internal policies, is necessary to maintain the /var partition mounted with the noexec option, besides ensuring that /var/lib/docker or alternative Docker root directory mount excludes the noexec option, an additional partition must be created for the /var/lib/replicated-operator mount with a minimum of 2GB and also exempt from using the aforementioned option.

Outcome

The Terraform Enterprise application will now start successfully. Watch the containers being launched and check the logs:

Watch the docker containers being launched
```
watch docker ps -f network=tfe_services
```

Check the previously failed containers logs to confirm the issue has been resolved, example:

$ docker logs -f tfe-vault
...
+ Configuring Vault
+ mounting vault transit backend
+ tuning vault lease TTLs
Success! Data written to: sys/mounts/auth/token/tune
+ adding vault policy
Success! Uploaded policy: atlas
+ Creating vault token for use by services
+ Successfully created vault token

Additional Information