feat(template): load coder agent env vars from k8s secret if available

templates/kubernetes/homelab-workspace/deployment.tf

@@ -54,10 +54,6 @@ resource "kubernetes_deployment_v1" "deployment" {
             name  = "HOMEBREW_PREFIX"
             value = local.homebrew_directory
           }
-          env {
-            name  = "NORMALISE_TMP_PERMISSIONS"
-            value = "true"
-          }
           volume_mount {
             mount_path = local.home_directory
             name       = "home"
@@ -84,19 +80,19 @@ resource "kubernetes_deployment_v1" "deployment" {
           }
         }
         container {
-          name = "workspace"
-          # Not Coder's generated /scripts/workspace-init.sh directly: the entrypoint
-          # wipes /tmp and then execs it. The wipe has to happen before the
-          # agent unpacks its CLI into /tmp, and this is the only hook that runs
-          # on a container-only restart within a live Pod (init containers do
-          # not) - see script-container-entrypoint.sh and the "tmp" volume
-          # below.
+          name    = "workspace"
           command = ["/bin/bash", "/scripts/script-container-entrypoint.sh"]
           image   = var.workspace_image
           env {
             name  = "CODER_AGENT_TOKEN"
             value = coder_agent.main.token
           }
+          env_from {
+            secret_ref {
+              name     = "coder-workspace-env"
+              optional = true
+            }
+          }
           liveness_probe {
             exec {
               command = ["/bin/sh", "-c", "pgrep -f \"coder agent\" || exit 1"]
@@ -270,12 +266,7 @@ resource "kubernetes_deployment_v1" "deployment" {
         # node notices. Its lifecycle matches the Pod's (created fresh, deleted
         # with it) - like the "system" volume above, that means a Pod restart
         # gets a clean volume but a container-only restart within a live Pod
-        # does not, which is why the container's entrypoint
-        # (script-container-entrypoint.sh) wipes /tmp's contents explicitly on
-        # every container start instead of relying on this. That wipe belongs in
-        # the entrypoint and nowhere later: the Coder agent unpacks its own CLI
-        # into /tmp before it runs anything else, so a wipe from the agent
-        # startup script deletes it.
+        # does not.
         volume {
           name = "tmp"
           ephemeral {

templates/kubernetes/homelab-workspace/scripts/script-agent-startup.sh

@@ -1,6 +1,9 @@
 #!/bin/bash
 set -eo pipefail
 
+ENV_VARS_FILE="/home/coder/.env.pod"
+
+
 # Coder's bootstrap unpacks the agent CLI into a per-boot directory under /tmp
 # and the agent appends that directory to the PATH of everything it runs, so
 # resolving "coder" here goes through exactly the same lookup a metadata script
@@ -32,6 +35,9 @@ main() {
     echo 'set -o allexport; source /etc/environment; set +o allexport' >> ~/.bashrc
     echo '------------------------------------------------------------'
   fi
+  if [[ -f "${ENV_VARS_FILE}" ]]; then
+    set -a; source "${ENV_VARS_FILE}"; set +a
+  fi
   echo 'Done'
 }
 

templates/kubernetes/homelab-workspace/scripts/script-container-entrypoint.sh

@@ -1,7 +1,23 @@
 #!/bin/bash
 set -eo pipefail
 
+ENV_VARS_FILE="/home/coder/.env.pod"
+
+
+record_env_vars() {
+  if [[ -f "${ENV_VARS_FILE}" ]]; then
+    rm -f "${ENV_VARS_FILE}"
+  fi
+  touch "${ENV_VARS_FILE}"
+  for key in $(env | cut -d= -f1 | grep '^CODER_VAR_'); do
+    value=$(printenv "$key")
+    echo "$key=\"$value\"" >> "${ENV_VARS_FILE}"
+  done
+}
+
 main() {
+  record_env_vars
+
   # Hand off to Coder's generated agent bootstrap, replacing this process rather
   # than spawning it: the agent has to stay PID 1, both because it reaps orphans
   # in this container and because the Deployment's liveness probe pgreps for it.

templates/kubernetes/homelab-workspace/scripts/script-prepare-workspace.sh

@@ -81,10 +81,6 @@ prepare_environment() {
 }
 
 normalise_tmp_permissions() {
-  if [[ "${NORMALISE_TMP_PERMISSIONS:-}" != "true" ]]; then
-    return
-  fi
-
   # fsGroup makes the production volume group-writable. The sticky bit keeps
   # processes that share /tmp from replacing one another's entries.
   chmod 1777 /tmp