deploy/uno-q/recover-uno-q.sh

@@ -36,7 +36,9 @@ echo "    forward :8080 · reverse :$SHIM_PORT"
 
 echo "==> [3/5] ensure on-board supervisor is running"
 if a shell "test -x $SUPERVISOR" 2>/dev/null; then
-  a shell "pgrep -f zeroclaw-supervisor >/dev/null 2>&1 \
+  # Bracket the pattern so the pgrep's own shell (whose args contain the
+  # supervisor path) doesn't self-match and skip the launch.
+  a shell "pgrep -f '[z]eroclaw-supervisor.sh' >/dev/null 2>&1 \
     || (setsid nohup $SUPERVISOR >/dev/null 2>&1 < /dev/null &)" >/dev/null 2>&1
   echo "    supervisor ensured (restarts llama + daemon on death/wedge)"
 else

deploy/uno-q/zeroclaw-supervisor.sh

@@ -16,6 +16,9 @@
 # Boot persist: crontab -l | { cat; echo '@reboot /home/arduino/zeroclaw-supervisor.sh'; } | crontab -
 set -u
 
+# cron's @reboot env is minimal — guarantee the tools we shell out to are found.
+export PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:${PATH:-}
+
 HOME_DIR=/home/arduino
 LLAMA_DIR="$HOME_DIR/llama"
 MODEL="$HOME_DIR/models/qwen.gguf"
@@ -30,11 +33,17 @@ log() { echo "[$(date '+%F %T')] $*" >> "$LOG"; }
 now() { date +%s; }
 healthy() { curl -sf --max-time 4 "http://127.0.0.1:$1/health" >/dev/null 2>&1; }
 
-# single-instance guard (no flock dependency)
+# single-instance guard (no flock dependency). Pid-reuse-safe: a stale lock
-if [ -f "$LOCK" ] && kill -0 "$(cat "$LOCK" 2>/dev/null)" 2>/dev/null; then
+# whose pid was recycled by an unrelated process (common right after a reboot)
-  log "supervisor already running (pid $(cat "$LOCK")) — exiting"
+# must NOT block startup — so require the live pid to actually be a supervisor.
+if [ -f "$LOCK" ]; then
+  OLDPID=$(cat "$LOCK" 2>/dev/null)
+  if [ -n "$OLDPID" ] && kill -0 "$OLDPID" 2>/dev/null \
+     && grep -qa zeroclaw-supervisor "/proc/$OLDPID/cmdline" 2>/dev/null; then
+    log "supervisor already running (pid $OLDPID) — exiting"
     exit 0
   fi
+fi
 echo $$ > "$LOCK"
 trap 'rm -f "$LOCK"' EXIT