detect starting wan, plus tons of bug fixes and overall improvements close #1

failover.sh

@@ -1,26 +1,55 @@
 #!/bin/bash
+#wan failover script. ping multiple hosts and failover from primary to secondary after too many fails on primary
 
+#how often to send pings
 INTERVAL=1
+#how many packets to send to each host
 PACKETS=1
+#how often (in seconds) to re-check primary wan while on secondary wan and primary had failed
+RECOVERY_CHECK=300
 #reliable hosts for checking are cloudflare, quad9, hgoogle
 HOSTS=("1.1.1.1" "9.9.9.9" "8.8.8.8")
-WAN1=eth0
-WAN2=wwan0
-USINGWAN=eth0
-CHECKWAN=eth0
-COUNTER=0
-CHANGED=0
+#interface and luci names of primary and secondary wan
+WAN1_IFACE_NAME="eth0"
+WAN2_IFACE_NAME="wwan0"
+WAN1_LUCI_NAME="wan"
+WAN2_LUCI_NAME="wanb" #currently unused
 
+#get the current active wan interface, in case the script is started during failover
+function get_active_wan {
+	local ROUTE=0
+	local CURRENT_METRIC=""
+	local ROUTE_IFACE=""
+	while [ $ROUTE -le 100 ]
+	do
+		ROUTE_IFACE=`uci get network.@route[$ROUTE].interface`
+		if [ "$ROUTE_IFACE" == "uci: Entry not found" ]; then
+			break
+		elif [ "$ROUTE_IFACE" == "$WAN1_LUCI_NAME" ]; then
+			CURRENT_METRIC=`uci get network.@route[$ROUTE].metric`
+			break
+		fi
+		let ROUTE++
+	done
+	if [ "$CURRENT_METRIC" == "0" ]; then
+		#current wan1 metric is 0, so it is set as primary
+		echo $WAN1_IFACE_NAME
+	else
+		echo $WAN2_IFACE_NAME
+	fi
+}
+
+#change the primary wan metric
 function change_wan_metric() {
 	CHANGED=$SECONDS
 	local ROUTE=0
-	# 100 just in case this runs away for some reason
+	local ROUTE_IFACE=""
 	while [ $ROUTE -le 100 ]
 	do
-		local ROUTE_IFACE=`uci get network.@route[$ROUTE].interface`
-		if [ "$ROUTE_INTERFACE" == "uci: Entry not found" ]; then
+		ROUTE_IFACE=`uci get network.@route[$ROUTE].interface`
+		if [ "$ROUTE_IFACE" == "uci: Entry not found" ]; then
 			break
-		elif [ "$ROUTE_INTERFACE" == "wan" ]; then
+		elif [ "$ROUTE_IFACE" == "$WAN1_LUCI_NAME" ]; then
 			uci set network.@route[$ROUTE].metric='$1'
 			uci commit
 			reload_config
@@ -28,22 +57,31 @@ function change_wan_metric() {
 		fi
 		let ROUTE++
 	done
+	logger -t failover "`date`: Changed active WAN to $USINGWAN"
 }
 
 function debug {
 	logger -t failover "`date`: pings:$PINGS, counter:$COUNTER, changed:$CHANGED, seconds:$SECONDS, usingwan:$USINGWAN, checkwan:$CHECKWAN"
 }
 
-logger -t failover "`date`: Failover script started."
+#which wan is currently being used/checked
+USINGWAN=$(get_active_wan)
+CHECKWAN=$(get_active_wan)
+#how many times all pings failed
+COUNTER=0
+#when wan was last changed
+CHANGED=0
+
+logger -t failover "`date`: Failover script started, current wan $USINGWAN"
 while sleep $INTERVAL
 do
 
-	if [ "$USINGWAN" == "$WAN2" ] && [ $(($SECONDS-$CHANGED)) -gt 300 ]; then
+	if [ "$USINGWAN" == "$WAN2_IFACE_NAME" ] && [ $(($SECONDS-$CHANGED)) -gt $RECOVERY_CHECK ]; then
 		debug
-		# after 5 minutes on failover wan2, check main wan. if it fails counter will go to 1 but since
-		# wan2 is still fine it should reset to 0 next time and stay on wan2, until the check repeats and so on.
-		CHECKWAN=$WAN1
-		CHANGED=$SECONDS #and reset changed so it just tried once then returns to checking working wan
+		CHECKWAN=$WAN1_IFACE_NAME
+		CHANGED=$SECONDS
+		#NOTE: we dont do a modulo check because and instead just reset CHANGED because pings can take longer than the interval
+		#during a failure so it can potentially take multiples of RECOVERY_CHECK even after primary wan has recovered
 	fi
 
 	PINGS=0
@@ -55,12 +93,11 @@ do
 			fi
 	done
 
-	#debug if any pings fail, just for fun
+	#debug if any pings fail
 	if [ $PINGS -lt ${#HOSTS[@]} ]; then
 		debug
 	fi
 
-
 	# if all pings failed increase counter
 	if [ $PINGS -eq 0 ]; then
 		let COUNTER++
@@ -69,31 +106,24 @@ do
 	fi
 
 	# counter failed pings to all hosts multiple times, switch WAN
+	# NOTE: if both wan fail somultaneously, it will just switch back and forth. but not like it matters, neither would work anyway... lol
 	if [ $COUNTER -gt 3 ]; then
 		debug
-		# NOTE: if both wan fail, it will just switch back and forth. but not like it matters, neither would work anyway... lol
-		if [ "$USINGWAN" == "$WAN1" ]; then
+		if [ "$USINGWAN" == "$WAN1_IFACE_NAME" ]; then
 			#if failed and currently on wan, switch to wanb
+			USINGWAN=$WAN2_IFACE_NAME
 			change_wan_metric 2
-			USINGWAN=$WAN2
-			CHECKWAN=$USINGWAN
-			logger -t failover "`date`: Changed active WAN metric to 4G modem!"
-		elif [ "$USINGWAN" == "$WAN2" ]; then
+		elif [ "$USINGWAN" == "$WAN2_IFACE_NAME" ]; then
 			#if failed and currently on wanb, switch back to wan
+			USINGWAN=$WAN1_IFACE_NAME
 			change_wan_metric 0
-			USINGWAN=$WAN1
-			CHECKWAN=$USINGWAN
-			logger -t failover "`date`: Changed active WAN metric to Cable connection!"
 		fi
 	elif [ $COUNTER -eq 0 ]; then
-	# counter is successful, if on wan2 and wan1 has recovered return to wan1
-		if [ "$CHECKWAN" == "$WAN1" ] && [ "$USINGWAN" == "$WAN2" ]; then
+		if [ "$CHECKWAN" == "$WAN1_IFACE_NAME" ] && [ "$USINGWAN" == "$WAN2_IFACE_NAME" ]; then
+			#ping(s) successful, if on wan2 and wan1 has recovered return to wan1
 			debug
-			#if failed and currently on wanb, switch back to wan
+			USINGWAN=$WAN1_IFACE_NAME
 			change_wan_metric 0
-			USINGWAN=$WAN1
-			CHECKWAN=$USINGWAN
-			logger -t failover "`date`: Changed active WAN metric to Cable connection!"
 		fi
 	fi