From 41c0598220ef6427123cf643841ee97dff6bae7c Mon Sep 17 00:00:00 2001
From: Gmer4Lfe <gmer4lfe@gmail.com>
Date: Sat, 22 Aug 2026 00:00:56 -0400
Subject: [PATCH] A fallback dry run wrote real reboot-surviving state, and a
 failed tier writeback reported itself as a dry run

---
 Fallback/fallback.sh      | 28 ++++++++++++++++++++++++++--
 Fallback/fallback_test.sh | 21 +++++++++++++++++++++
 2 files changed, 47 insertions(+), 2 deletions(-)

diff --git a/Fallback/fallback.sh b/Fallback/fallback.sh
index 15cf5ba..eebec5e 100755
--- a/Fallback/fallback.sh
+++ b/Fallback/fallback.sh
@@ -368,6 +368,22 @@ state_set() {
 }
 
 state_init() {
+    # A dry run must not leave the host believing it failed over. state_set() writes
+    # unconditionally, and this file survives reboots and is what the real daemon — and the
+    # Monitor and Fallback cards — read to decide what is happening. A --dry-run walk through
+    # FAILOVER would have written state=FALLBACK, the tier flags and the strike counter into it
+    # for real, and nothing would have put them back.
+    #
+    # Copied rather than merely redirected, so the preview still starts from the live state and
+    # can advance through tiers exactly as a real run would. The copy lands in the RAM cache and
+    # dies with the reboot.
+    if [[ "$DRY_RUN" == true ]]; then
+        local live="$FALLBACK_STATE_FILE"
+        FALLBACK_STATE_FILE="${VV_CACHE_ROOT:-/tmp/varaverk}/fallback_state.dryrun.$$"
+        mkdir -p "$(dirname "$FALLBACK_STATE_FILE")"
+        if [[ -f "$live" ]]; then cp -f "$live" "$FALLBACK_STATE_FILE"; else : > "$FALLBACK_STATE_FILE"; fi
+        warn "DRY RUN — state writes redirected to $FALLBACK_STATE_FILE (live state untouched)"
+    fi
     mkdir -p "$(dirname "$FALLBACK_STATE_FILE")"
     [[ ! -f "$FALLBACK_STATE_FILE" ]] && touch "$FALLBACK_STATE_FILE"
     [[ -z "$(state_get state)" ]]                  && state_set state "NORMAL"
@@ -759,8 +775,16 @@ run_handback() {
                 for job in "${jobs[@]}"; do
                     [[ -z "$job" ]] && continue
                     log "Syncing: $job"
-                    [[ "$DRY_RUN" == false ]] && bash "$SCRIPT_DIR/../Rsync/rsync.sh" "$job" \
-                        || warn "DRY RUN — would rsync: $job"
+                    # if/else, not A && B || C. In the shorthand a REAL run whose rsync exits
+                    # non-zero falls through to the || branch and logs "DRY RUN — would rsync",
+                    # so a failed Tier writeback reported itself as a preview and the real
+                    # failure went unsaid. The Tier 1 block below always had this right.
+                    if [[ "$DRY_RUN" == false ]]; then
+                        bash "$SCRIPT_DIR/../Rsync/rsync.sh" "$job" \
+                            || error "Tier $tier writeback FAILED: $job"
+                    else
+                        warn "DRY RUN — would rsync: $job"
+                    fi
                 done
             else
                 log "Tier $tier writeback skipped — outage ${outage_minutes}min < ${threshold}min"
diff --git a/Fallback/fallback_test.sh b/Fallback/fallback_test.sh
index 30c070f..cb5d9f3 100755
--- a/Fallback/fallback_test.sh
+++ b/Fallback/fallback_test.sh
@@ -325,6 +325,27 @@ else
     warn "No state file found — assuming NORMAL (first run)"
 fi
 
+# fallback.sh must actually be RUNNING, not merely enabled
+#
+# Every phase after this one waits for the daemon to change state. FALLBACK_ENABLED=true says
+# it is allowed to run; it does not say array_started.sh launched it, or that it is still alive.
+# Without this the test passes pre-flight, drops a real iptables rule on the partner, waits
+# FALLBACK_TEST_BLOCK_WAIT for a transition nothing is there to make, and fails Phase 3 blaming
+# fallback detection. Only the EXIT trap gets connectivity back.
+#
+# In --dry-run nothing is blocked and nothing is waited on, so a dead daemon is worth saying but
+# not worth aborting for — the walkthrough still shows the operator the shape of the run.
+if pgrep -f "Fallback/fallback\.sh" >/dev/null 2>&1; then
+    log "fallback.sh daemon is running"
+elif [[ "$DRY_RUN" == true ]]; then
+    warn "fallback.sh is NOT running — a real test would abort here"
+else
+    error "fallback.sh is not running — nothing would detect the outage this test creates"
+    error "Start it with array_started.sh, or run with --dry-run to walk the phases"
+    phase_fail "Pre-flight"
+    exit 1
+fi
+
 # Tier 1 containers configured
 if [[ ${#TIER1_CONTAINERS[@]} -eq 0 ]]; then
     error "No Tier 1 containers configured for $MY_ID → $REMOTE_ID"