router: per-list tables, gateway metric, per-gateway failover

router/functions

@@ -25,6 +25,16 @@ table_by_name() { awk -v n="$1" '$2 == n { print $1; exit }' /etc/iproute2/rt_ta
 # Resolve "default" keyword to actual default gateway
 resolve_default_gw() { $1 route show default | awk '/default/ {print $3; exit}' ; }
 
+# Parse gateway line: "IP [metric N]" → sets gw_ip, gw_metric
+# Usage: parse_gw_line "91.232.225.122 metric 10" "ip"
+parse_gw_line()
+{
+    local line="$1" ipcmd="${2:-ip}"
+    local raw=$(echo "$line" | awk '{print $1}')
+    gw_ip=$(resolve_gw "$raw" "$ipcmd")
+    gw_metric=$(echo "$line" | awk '/metric/{for(i=1;i<=NF;i++) if($i=="metric") print $(i+1)}')
+}
+
 # Resolve gateway value: "default" → system default, hostname → IP, IP → as-is
 resolve_gw()
 {

router/route-health.sh

@@ -2,6 +2,7 @@
 # Health-check failover for route groups via Telegraf/InfluxDB
 # Monitors gateway health and flushes dead routes when live alternatives exist
 # Manages default route for groups with set-default option
+# Supports per-list tables and per-gateway failover (metric mode)
 #
 # Usage: route-health.sh [--show]
 
@@ -69,6 +70,51 @@ get_health()
     }'
 }
 
+# Find monitor tag for a gateway IP by scanning all groups
+# Usage: find_gw_monitor ROUTES_DIR GW_IP IPCMD
+find_gw_monitor()
+{
+    local routes_dir="$1" gw_ip="$2" ipcmd="$3"
+    for check_dir in "$routes_dir"/*/ ; do
+        [ -d "$check_dir" ] || continue
+        [ -f "$check_dir/monitor" ] || continue
+        local check_ip=$(resolve_gw "$(read_value "$check_dir/gateway")" "$ipcmd" 2>/dev/null)
+        if [ "$check_ip" = "$gw_ip" ] ; then
+            read_value "$check_dir/monitor"
+            return
+        fi
+    done
+}
+
+# Flush per-list tables for a group
+# Usage: flush_group_tables STATE_GROUP_DIR IPCMD
+flush_group_tables()
+{
+    local group_state="$1" ipcmd="$2"
+    for list_state in "$group_state"/*/ ; do
+        [ -d "$list_state" ] || continue
+        [ -f "$list_state/table" ] || continue
+        local t ; read -r t < "$list_state/table"
+        $ipcmd route flush table "$t" 2>/dev/null
+    done
+}
+
+# Remove routes via specific gateway from all per-list tables
+# Usage: flush_gw_routes STATE_GROUP_DIR GW_IP IPCMD
+flush_gw_routes()
+{
+    local group_state="$1" dead_gw="$2" ipcmd="$3"
+    for list_state in "$group_state"/*/ ; do
+        [ -d "$list_state" ] || continue
+        [ -f "$list_state/table" ] || continue
+        local t ; read -r t < "$list_state/table"
+        # Remove routes via dead gateway
+        $ipcmd route show table "$t" 2>/dev/null | grep "via $dead_gw" | while read -r route ; do
+            $ipcmd route del $route table "$t" 2>/dev/null
+        done
+    done
+}
+
 # --- Evaluate health for all monitored groups ---
 healthy_count=0
 dead_count=0
@@ -129,24 +175,12 @@ for routes_dir in "$ROUTES_DIR" "$ROUTES6_DIR" ; do
             [ -z "$line" ] && continue
             echo "$line" | grep -q '^#' && continue
 
-            gw_ip=$(resolve_gw "$line" "$ipcmd")
+            parse_gw_line "$line" "$ipcmd"
             [ -z "$gw_ip" ] && continue
 
-            # Find monitor tag for this gateway IP
-            # Look through all groups for one whose gateway matches
-            gw_tag=""
-            for check_dir in "$routes_dir"/*/ ; do
-                [ -d "$check_dir" ] || continue
-                [ -f "$check_dir/monitor" ] || continue
-                check_ip=$(resolve_gw "$(read_value "$check_dir/gateway")" "$ipcmd" 2>/dev/null)
-                if [ "$check_ip" = "$gw_ip" ] ; then
-                    gw_tag=$(read_value "$check_dir/monitor")
-                    break
-                fi
-            done
+            gw_tag=$(find_gw_monitor "$routes_dir" "$gw_ip" "$ipcmd")
 
             if [ -z "$gw_tag" ] ; then
-                # No monitor for this gateway — assume healthy (first unmonitored wins)
                 [ -n "$SHOW" ] && log "[$name] $gw_ip: no monitor, assuming healthy"
                 best_gw="$gw_ip"
                 break
@@ -182,16 +216,11 @@ done
 for state_path in $dead_groups ; do
     name=$(basename "$state_path")
     routes_dir=$(dirname "$state_path")
+    gwdir="$state_path"
 
     mkdir -p "$STATE_DIR/$state_path"
 
-    # Read table number from group config or rt_tables
-    table=""
-    if [ -f "$state_path/table" ] ; then
-        table=$(read_value "$state_path/table")
-    else
-        table=$(table_by_name "$name")
-    fi
+    ipcmd=$(ipcmd_for "$routes_dir")
 
     # Flapping protection: require DOWN_THRESHOLD consecutive checks
     down_count=0
@@ -217,11 +246,44 @@ for state_path in $dead_groups ; do
         continue
     fi
 
-    ipcmd=$(ipcmd_for "$routes_dir")
+    # Check if this is a metric group with per-gateway failover
+    has_metric=""
+    [ -f "$gwdir/gateway" ] && grep -v '^#' "$gwdir/gateway" | grep -q 'metric' && has_metric=1
+
+    if [ -n "$has_metric" ] ; then
+        # Metric mode: only remove routes via this dead gateway, keep others
+        # Find dead gateway IP from monitor tag
+        tag=$(read_value "$gwdir/monitor")
+        dead_gw=""
+        while IFS= read -r gw_line ; do
+            [ -z "$gw_line" ] && continue
+            echo "$gw_line" | grep -q '^#' && continue
+            parse_gw_line "$gw_line" "$ipcmd"
+            [ -z "$gw_ip" ] && continue
+            gw_tag=$(find_gw_monitor "$routes_dir" "$gw_ip" "$ipcmd")
+            if [ "$gw_tag" = "$tag" ] ; then
+                dead_gw="$gw_ip"
+                break
+            fi
+        done < "$gwdir/gateway"
+
+        if [ -n "$dead_gw" ] ; then
+            log "[$name] Dead (${down_count}x confirmed), removing routes via $dead_gw (metric failover)"
+            if [ -z "$SHOW" ] ; then
+                flush_gw_routes "$STATE_DIR/$state_path" "$dead_gw" "$ipcmd"
+            fi
+        else
+            log "[$name] Dead (${down_count}x confirmed), cannot identify dead gateway IP"
+        fi
+    else
+        # Non-metric: flush all per-list tables
+        log "[$name] Dead (${down_count}x confirmed), flushing all per-list tables"
+        if [ -z "$SHOW" ] ; then
+            flush_group_tables "$STATE_DIR/$state_path" "$ipcmd"
+        fi
+    fi
 
-    log "[$name] Dead (${down_count}x confirmed), flushing table $table"
-    if [ -z "$SHOW" ] && [ -n "$table" ] ; then
-        $ipcmd route flush table "$table" 2>/dev/null
+    if [ -z "$SHOW" ] ; then
         echo "down" > "$STATE_DIR/$state_path/health"
         echo "$down_count" > "$STATE_DIR/$state_path/down_count"
     fi
@@ -242,8 +304,11 @@ for state_path in $healthy_groups ; do
     log "[$name] Recovered, scheduling route reload"
     if [ -z "$SHOW" ] ; then
         echo "up" > "$STATE_DIR/$state_path/health"
-        # Remove route-update state to trigger reload on next run
-        rm -f "$STATE_DIR/$state_path/hash" "$STATE_DIR/$state_path/resolved"
+        # Remove per-list state hashes to trigger reload on next route-update run
+        for list_state in "$STATE_DIR/$state_path"/*/ ; do
+            [ -d "$list_state" ] || continue
+            rm -f "$list_state/hash" "$list_state/resolved"
+        done
         need_reload=1
     fi
 done