only heal the sonic db after a crash.

sonic/scripts/entrypoint.sh

@@ -1,6 +1,66 @@
 #!/bin/bash
 
 datadir=${SONIC_HOME:-/var/sonic}
+crash_loop_file="$datadir/.crash_loop_timestamps"
+crash_loop_threshold=3  # Number of restarts to consider it a crash loop
+crash_loop_window=60    # Time window in seconds
+
+# Function to detect crash loop
+detect_crash_loop() {
+    local current_time=$(date +%s)
+    
+    # Read existing timestamps and filter out old ones (outside the window)
+    if [ -f "$crash_loop_file" ]; then
+        local recent_restarts=0
+        while IFS= read -r timestamp; do
+            local age=$((current_time - timestamp))
+            if [ $age -lt $crash_loop_window ]; then
+                recent_restarts=$((recent_restarts + 1))
+            fi
+        done < "$crash_loop_file"
+        
+        # If we have enough recent restarts, it's a crash loop
+        if [ $recent_restarts -ge $((crash_loop_threshold - 1)) ]; then
+            return 0  # Crash loop detected
+        fi
+    fi
+    
+    return 1  # No crash loop
+}
+
+# Function to record restart timestamp
+record_restart() {
+    local current_time=$(date +%s)
+    local cutoff_time=$((current_time - crash_loop_window))
+    
+    # Filter out old timestamps and check if any recent ones exist
+    local temp_file=$(mktemp)
+    local has_recent=false
+    
+    if [ -f "$crash_loop_file" ]; then
+        while IFS= read -r timestamp; do
+            if [ "$timestamp" -gt "$cutoff_time" ]; then
+                echo "$timestamp" >> "$temp_file"
+                has_recent=true
+            fi
+        done < "$crash_loop_file" 2>/dev/null || true
+    fi
+    
+    # If no recent timestamps, previous run was successful - start fresh
+    # Otherwise, keep the recent timestamps and add the new one
+    if [ "$has_recent" = false ]; then
+        echo "$current_time" > "$temp_file"
+    else
+        echo "$current_time" >> "$temp_file"
+    fi
+    
+    mv "$temp_file" "$crash_loop_file" 2>/dev/null || true
+}
+
+# Function to clear crash loop tracking (called on successful startup)
+clear_crash_loop_tracking() {
+    rm -f "$crash_loop_file"
+}
 
 if [ ! -f "$datadir/initialized" ]; then
     echo "Initializing Sonic..."
@@ -14,18 +74,27 @@ if [ ! -f "$datadir/initialized" ]; then
     rm "$datadir/$filename"
     
     touch "$datadir/initialized"
+    clear_crash_loop_tracking
 
     echo "Initialization complete."
 else
     echo "Sonic is already initialized."
-    # just in case because after every shutdown this shit is corrupted by default
-    # GOMEMLIMIT=28GiB exec sonictool --datadir "$datadir" --cache 12000 heal
+    
+    # Record this restart attempt
+    record_restart
+    
+    # Check if we're in a crash loop
+    if detect_crash_loop; then
+        echo "Crash loop detected. Running database heal..."
+        # Use exec so it runs as PID 0 and can be interrupted by docker
+        # After heal completes, we'll continue to start sonicd
+        GOMEMLIMIT="${CACHE_GB}GiB" sonictool --datadir "$datadir" --cache "${CACHE_GB}000" heal
+        echo "Heal completed. Starting sonicd..."
+    else
+        echo "No crash loop detected. Skipping heal."
+    fi
 fi
 
-# after every restart I needed to do this manually so lets just do it automatically each time
-
-GOMEMLIMIT="${CACHE_GB}GiB" sonictool --datadir "$datadir" --cache "${CACHE_GB}000" heal
-
 #echo "Generating new Geth node key..."
 #openssl rand 32 | xxd -p -c 32 | tr -d '\n' > "$datadir/nodekey"