From eee6ce5581c72db5df0c4c9428575f0ca53ab052 Mon Sep 17 00:00:00 2001
From: CallMeTechie <34693633+CallMeTechie@users.noreply.github.com>
Date: Wed, 24 Jun 2026 07:42:23 +0200
Subject: [PATCH] chore(deploy): prune dangling images after a successful
 auto-update
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Each :latest pull leaves the previous gatecontrol image dangling. update.sh never
removed them, so the Docker host filled up over time — incident 2026-06-23: 187
images, root fs at 100%, 'docker pull' failed with 'no space left on device', and
the */5 auto-update silently stalled on the old version.

Add a prune_images() helper (docker image prune -f — dangling only, tagged/in-use
images are never touched) and call it after a successful recreate in both the
manual and auto paths. Applied to the served template and the repo copy; the live
/opt/gatecontrol/update.sh was synced to this version too (it had also drifted —
it was missing the ensure_guacd sidecar step).
---
 src/services/systemSetup/templates/update.sh | 12 ++++++++++--
 update.sh                                    | 12 ++++++++++--
 2 files changed, 20 insertions(+), 4 deletions(-)

diff --git a/src/services/systemSetup/templates/update.sh b/src/services/systemSetup/templates/update.sh
index 4ddc7441..f760199c 100755
--- a/src/services/systemSetup/templates/update.sh
+++ b/src/services/systemSetup/templates/update.sh
@@ -97,6 +97,14 @@ ensure_guacd() {
   )
 }
 
+prune_images() {
+  # Reclaim disk after a successful update: each :latest pull leaves the previous
+  # image dangling. Without pruning the host fills up over time (incident
+  # 2026-06-23: 187 images → root fs 100% → pulls failed → auto-update stalled).
+  # -f removes ONLY dangling (untagged) images, so tagged/in-use images are safe.
+  docker image prune -f >>"$LOG" 2>&1 || true
+}
+
 needs_update() { # echoes "yes" if running image != :latest
   local latest running
   latest="$(docker image inspect "$IMAGE" --format '{{.Id}}' 2>/dev/null || true)"
@@ -116,7 +124,7 @@ if [ "$MODE" = "manual" ]; then
     err) log "could not resolve :latest digest"; write_state failed manual; exit 1 ;;
   esac
   log "manual trigger — recreating"
-  if recreate; then ensure_guacd || true; write_state updated manual; else log "recreate/health failed"; write_state failed manual; exit 1; fi
+  if recreate; then ensure_guacd || true; prune_images || true; write_state updated manual; else log "recreate/health failed"; write_state failed manual; exit 1; fi
   exit 0
 fi
 
@@ -128,4 +136,4 @@ case "$(needs_update)" in
   err) log "could not resolve :latest digest"; write_state failed auto; exit 1 ;;
 esac
 log "auto: update needed — recreating"
-if recreate; then ensure_guacd || true; write_state updated auto; else log "recreate/health failed"; write_state failed auto; exit 1; fi
+if recreate; then ensure_guacd || true; prune_images || true; write_state updated auto; else log "recreate/health failed"; write_state failed auto; exit 1; fi
diff --git a/update.sh b/update.sh
index 4ddc7441..f760199c 100755
--- a/update.sh
+++ b/update.sh
@@ -97,6 +97,14 @@ ensure_guacd() {
   )
 }
 
+prune_images() {
+  # Reclaim disk after a successful update: each :latest pull leaves the previous
+  # image dangling. Without pruning the host fills up over time (incident
+  # 2026-06-23: 187 images → root fs 100% → pulls failed → auto-update stalled).
+  # -f removes ONLY dangling (untagged) images, so tagged/in-use images are safe.
+  docker image prune -f >>"$LOG" 2>&1 || true
+}
+
 needs_update() { # echoes "yes" if running image != :latest
   local latest running
   latest="$(docker image inspect "$IMAGE" --format '{{.Id}}' 2>/dev/null || true)"
@@ -116,7 +124,7 @@ if [ "$MODE" = "manual" ]; then
     err) log "could not resolve :latest digest"; write_state failed manual; exit 1 ;;
   esac
   log "manual trigger — recreating"
-  if recreate; then ensure_guacd || true; write_state updated manual; else log "recreate/health failed"; write_state failed manual; exit 1; fi
+  if recreate; then ensure_guacd || true; prune_images || true; write_state updated manual; else log "recreate/health failed"; write_state failed manual; exit 1; fi
   exit 0
 fi
 
@@ -128,4 +136,4 @@ case "$(needs_update)" in
   err) log "could not resolve :latest digest"; write_state failed auto; exit 1 ;;
 esac
 log "auto: update needed — recreating"
-if recreate; then ensure_guacd || true; write_state updated auto; else log "recreate/health failed"; write_state failed auto; exit 1; fi
+if recreate; then ensure_guacd || true; prune_images || true; write_state updated auto; else log "recreate/health failed"; write_state failed auto; exit 1; fi