From 7b7d9965b9069078544f87dc96c97b6b66a8294f Mon Sep 17 00:00:00 2001
From: Simon Marsh <simon@burble.com>
Date: Sat, 29 Jun 2019 10:52:40 +0100
Subject: [PATCH] Refactor SOA checking, parallel DNS queries

---
 contrib/#dn42promsrv.service# |  29 ----------
 dns.go                        | 101 +++++++++++++++++++++++++---------
 2 files changed, 76 insertions(+), 54 deletions(-)
 delete mode 100644 contrib/#dn42promsrv.service#

diff --git a/contrib/#dn42promsrv.service# b/contrib/#dn42promsrv.service#
deleted file mode 100644
index 0349925..0000000
--- a/contrib/#dn42promsrv.service#
+++ /dev/null
@@ -1,29 +0,0 @@
-##########################################################################
-# dn42promsrv example systemd service file
-##########################################################################
-
-[Unit]
-Description=DN42 Prometheus Stats Server
-After=network.target
-
-[Install]
-WantedBy=multi-user.target
-
-[Service]
-User=promsrv
-Group=promsrv
-Type=simple
-Restart=on-failure
-# service hardening
-ProtectSystem=strict
-NoNewPrivileges=yes
-ProtectControlGroups=yes
-PrivateTmp=yes
-PrivateDevices=yes
-DevicePolicy=closed
-MemoryDenyWriteExecute=yes
-#
-ExecStart=/usr/local/bin/dn42promsrv 
-     
-#########################################################################
-# end of file
diff --git a/dns.go b/dns.go
index 0516437..9daf3d2 100644
--- a/dns.go
+++ b/dns.go
@@ -18,6 +18,7 @@ import (
 	"net/http"
 	"strconv"
 	"strings"
+	"sync"
 	"time"
 )
 
@@ -174,7 +175,8 @@ func (m *DNSMetrics) Register() {
 
 	m.valid = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Name: "dn42_dns_valid",
-		Help: "0 = response and correct serial, 1 = response but incorrect serial, 2 = no response",
+		Help: "0 = response and latest serial, 1 = response and matching serial, " +
+			"2 = response but incorrect serial, 3 = server error",
 	}, []string{"role", "owner", "name", "ip", "addr"})
 	prometheus.MustRegister(m.valid)
 
@@ -204,44 +206,70 @@ func (m *DNSMetrics) Register() {
 
 func (m *DNSMetrics) Collect() {
 
-	now := uint64(time.Now().Unix())
-
-	// search the masters for the highest SOA
-	var latest_soa uint32 = 0
-
-	for _, server := range dns_servers {
-		if server.role == "master" {
-			if server.soa > latest_soa {
-				latest_soa = server.soa
-			}
-		}
-	}
-
-	// hold the results in an array where each entry corresponds
-	// to the dns_servers array
+	// hold the collect results in an array where each
+	// entry corresponds to the dns_servers array
 	results := make([]*DNSResult, len(dns_servers))
 
 	// query each server up to 3 times to try and get a result
 	for count := 0; count < 3; count++ {
 
+		// run each tranche of queries in parallel
+		var wg sync.WaitGroup
+
 		for ix := 0; ix < len(results); ix++ {
 			if results[ix] == nil {
 				// no result yet, query the server
-				results[ix] = dns_servers[ix].Query()
+
+				wg.Add(1)
+				go func(i int, s *sync.WaitGroup) {
+					results[i] = dns_servers[i].Query()
+					s.Done()
+				}(ix, &wg)
 			}
+
 		}
 
+		// all done
+		wg.Wait()
+
 	}
 
+	now := uint64(time.Now().Unix())
+
+	// create a list of master servers SOA and
+	// record which one has the latest version
+	var latest_soa uint32 = 0
+	masters := make([]uint32, 2)
+
+	for ix, server := range dns_servers {
+		if server.role == "master" {
+			soa := results[ix].serial
+			masters = append(masters, soa)
+			if soa > latest_soa {
+				latest_soa = soa
+			}
+		}
+	}
+
+	//	fmt.Printf("latest_soa %d\n", latest_soa)
+
 	// now go through each result and update the metrics
+	//
+	// valid = 0 - Serial number matches latest master
+	// valid = 1 - Serial number matches one master (but not the latest)
+	// valid = 2 - Serial number doesn't match anything
+	// valid = 3 - An error occured
+	//
+
 	for ix, r := range results {
 
-		var valid uint = 2
+		// assume that an error occured unless told otherwise
+		var valid uint = 3
 
 		// check if there was a valid result
 		if r != nil {
 
-			// update the server SOA
+			// update the server SOA from the result
 			dns_servers[ix].soa = r.serial
 
 			// SOA and RTT are direct metrics returned from the query
@@ -257,11 +285,32 @@ func (m *DNSMetrics) Collect() {
 				}).Info("DNS Server high RTT")
 			}
 
-			// check if the SOA matches the lastest master SOA
-			if r.serial == latest_soa {
-				valid = 0
-			} else {
-				valid = 1
+			// check if the SOA is valid
+
+			// it's an error if the SOA is more than 25 hours old
+			if (now - uint64(r.serial)) < (3600 * 25) {
+
+				//				fmt.Printf("Checking serial: %s = %d\n", dns_servers[ix].name, r.serial)
+
+				if r.serial == latest_soa {
+					// the SOA matches the current latest SOA
+					valid = 0
+
+				} else {
+
+					// assume no match found
+					valid = 2
+
+					// step through each master to see if the SOA matches somewhere
+					for _, soa := range masters {
+						if r.serial == soa {
+							// found one
+							valid = 1
+							break
+						}
+					}
+				}
+
 			}
 
 			// before setting whether the server is valid, calculate the stime
@@ -273,7 +322,7 @@ func (m *DNSMetrics) Collect() {
 
 			if uint64(r.serial) > now {
 				// server error
-				valid = 2
+				valid = 3
 			} else {
 				m.stime.With(m.label_map[ix]).Set(float64(now - uint64(r.serial)))
 			}
@@ -324,6 +373,8 @@ func (s *DNSServer) Query() *DNSResult {
 	client := new(dns.Client)
 	client.Timeout, _ = time.ParseDuration("4s")
 
+	//	fmt.Printf("Querying: %s\n", s.name)
+
 	// and finally query the server
 	resp, rtt, err := client.Exchange(msg, s.addr)
 	if err != nil || len(resp.Answer) != 1 {