prometheus: export count of unreachable peers (#3119)

* prometheus: export count of unreachable peers
* extracted part of printIPAMRanges

prog/weaver/http.go

@@ -34,23 +34,7 @@ var rootTemplate = template.New("root").Funcs(map[string]interface{}{
 	},
 	"printIPAMRanges": func(router weave.NetworkRouterStatus, status ipam.Status) string {
 		var buffer bytes.Buffer
-
-		type stats struct {
-			ips       uint32
-			nickname  string
-			reachable bool
-		}
-
-		peerStats := make(map[string]*stats)
-
-		for _, entry := range status.Entries {
-			s, found := peerStats[entry.Peer]
-			if !found {
-				s = &stats{nickname: entry.Nickname, reachable: entry.IsKnownPeer}
-				peerStats[entry.Peer] = s
-			}
-			s.ips += entry.Size
-		}
+		peerStats := summariseIpamStats(&status)
 
 		printOwned := func(name string, nickName string, info string, ips uint32) {
 			percentageRanges := float32(ips) * 100.0 / float32(status.RangeNumIPs)
@@ -116,6 +100,25 @@ var rootTemplate = template.New("root").Funcs(map[string]interface{}{
 	"trimSuffix": strings.TrimSuffix,
 })
 
+type ipamStats struct {
+	ips       uint32
+	nickname  string
+	reachable bool
+}
+
+func summariseIpamStats(status *ipam.Status) map[string]*ipamStats {
+	peerStats := make(map[string]*ipamStats)
+	for _, entry := range status.Entries {
+		s, found := peerStats[entry.Peer]
+		if !found {
+			s = &ipamStats{nickname: entry.Nickname, reachable: entry.IsKnownPeer}
+			peerStats[entry.Peer] = s
+		}
+		s.ips += entry.Size
+	}
+	return peerStats
+}
+
 func countDNSEntries(entries []nameserver.EntryStatus) int {
 	count := 0
 	for _, entry := range entries {

prog/weaver/metrics.go

@@ -82,6 +82,16 @@ var metrics = []metric{
 				ch <- intGauge(desc, metrics.Flows)
 			}
 		}},
+	{desc("weave_ipam_unreachable_count", "Number of unreachable peers."),
+		func(s WeaveStatus, desc *prometheus.Desc, ch chan<- prometheus.Metric) {
+			var count int
+			for _, entry := range summariseIpamStats(s.IPAM) {
+				if !entry.reachable {
+					count++
+				}
+			}
+			ch <- intGauge(desc, count)
+		}},
 	{desc("weave_ipam_pending_allocates", "Number of pending allocates."),
 		func(s WeaveStatus, desc *prometheus.Desc, ch chan<- prometheus.Metric) {
 			if s.IPAM != nil {

site/tasks/manage/metrics.md

@@ -20,6 +20,7 @@ exposed:
 * `weave_max_ips` - Size of IP address space used by allocator.
 * `weave_dns_entries` - Number of DNS entries.
 * `weave_flows` - Number of FastDP flows.
+* `weave_ipam_unreachable_coun` - Number of unreachable peers.
 * `weave_ipam_pending_allocates` - Number of pending allocates.
 * `weave_ipam_pending_claims` - Number of pending claims.