-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathmain.go
433 lines (378 loc) · 12.3 KB
/
main.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
package main
import (
"bytes"
"encoding/binary"
"flag"
"fmt"
"log"
"net"
"net/http"
"os"
"syscall"
"time"
configfile "github.com/benjojo/ixp-xping/config-file"
sockettimestamp "github.com/benjojo/ixp-xping/socket-timestamp"
"github.com/oxtoacart/bpool"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/promhttp"
)
var runtimeConfig *configfile.XPingConfig
var bufferPool = bpool.NewBytePool(512, 10000)
func main() {
config := configfile.XPingConfig{
ListenHost: "0.0.0.0",
ListenPortStart: 32768 - 32,
PollRateMS: 250,
Peers: []string{},
AllowedCIDRs: []string{},
PrometheusPort: 9150,
}
// Program Flags
configFileLocation := flag.String("cfg.path", "/etc/ixp-xping.yaml", "Where to find the YAML config")
lowerListenPort := flag.Int("probe.lower.port", int(config.ListenPortStart),
"The port number to start 16 UDP listeners from. The next 16 ports from the number provided should be free")
listenHost := flag.String("probe.listen", config.ListenHost, "The IP to listen on for probes")
flag.Parse()
f, err := os.Open(*configFileLocation)
if err == nil {
c, err := configfile.Parse(f)
if err == nil {
config = c
} else {
log.Printf("failed to parse: %v", err)
}
} else {
log.Printf("did not find a config file in %v", *configFileLocation)
}
config.LONAPAutoConfig()
runtimeConfig = &config
err = runSanityCheck()
if err != nil {
log.Fatalf("Sanity check failed: %v", err)
}
myOwnIPs := make(map[string]bool)
// Start the pool of UDP listeners
// if the user is "lazy" and doing bind to *, then we will look at all interfaces and
// listen on each one. else we can do what they say
if runtimeConfig.ListenHost == "0.0.0.0" {
interfaces, err := net.Interfaces()
if err != nil {
log.Fatalf("Could not list network interfaces, cannot intelligently bind() on ports: %v", err)
}
for _, iface := range interfaces {
addrs, err := iface.Addrs()
if err != nil {
continue
}
for _, v := range addrs {
interfaceIP, ok := v.(*net.IPNet)
if ok {
myOwnIPs[interfaceIP.IP.String()] = true
// Only listen on CIDRs that we can likely reply on anyway...
if runtimeConfig.IsAllowedCIDR(interfaceIP.IP) {
for i := 0; i < totalFlowsPerPeer; i++ {
target := net.JoinHostPort(interfaceIP.IP.String(), fmt.Sprint(*lowerListenPort+i))
uaddr, err := net.ResolveUDPAddr("udp", target)
if err != nil {
log.Fatalf("Could not resolve to listen on %s - %v", target, err)
}
udpl, err := net.ListenUDP("udp", uaddr)
if err != nil {
log.Fatalf("Could not listen on %s - %v", target, err)
}
rawFile, _ := udpl.File()
err = syscall.BindToDevice(int(rawFile.Fd()), iface.Name)
if err != nil {
log.Printf("Unable to call BindToDevice, the replies may not be sent from the right interface!")
}
go startUDPProbeReplier(udpl)
}
}
}
}
}
} else {
for i := 0; i < 16; i++ {
target := net.JoinHostPort(*listenHost, fmt.Sprint(*lowerListenPort+i))
uaddr, err := net.ResolveUDPAddr("udp", target)
if err != nil {
log.Fatalf("Could not resolve to listen on %s - %v", target, err)
}
udpl, err := net.ListenUDP("udp", uaddr)
if err != nil {
log.Fatalf("Could not listen on %s - %v", target, err)
}
go startUDPProbeReplier(udpl)
}
}
// Now start reading our own config to see who we should be probing
for _, srcpeer := range config.Peers {
portAlloc := 1
if myOwnIPs[srcpeer] {
for _, targetPeer := range config.Peers {
if myOwnIPs[targetPeer] {
log.Printf("Refusing to send probes to myself (as I was configured to so)")
continue
}
for i := 0; i < totalFlowsPerPeer; i++ {
portAlloc++
go sendPeerProbes(
srcpeer,
net.JoinHostPort(targetPeer, fmt.Sprint(config.ListenPortStart+uint32(i))),
time.Duration(config.PollRateMS)*time.Millisecond,
// this ensures that each prober has it's own, static source port, so that metrics can
// consistent (hopefully) between xping reboots.
config.ListenPortStart+totalFlowsPerPeer+uint32(portAlloc),
)
time.Sleep(time.Millisecond)
}
}
}
}
http.Handle("/metrics", promhttp.Handler())
log.Fatalf("failed to listen on HTTP: %v",
http.ListenAndServe(fmt.Sprintf("0.0.0.0:%d", config.PrometheusPort), nil))
}
const (
totalFlowsPerPeer = 16
lossTrackWindow = time.Second * 10
metricUpdateTicker = time.Second * 5
)
type peerProbeReply struct {
buf []byte
arrived time.Time
}
func sendPeerProbes(sendingIP, peer string, pollRate time.Duration, localBindPort uint32) {
poll := time.NewTicker(pollRate)
updateMetricsTick := time.NewTicker(metricUpdateTicker)
localListentarget := net.JoinHostPort(sendingIP, fmt.Sprint(localBindPort))
sendingAddr, err := net.ResolveUDPAddr("udp", localListentarget)
if err != nil {
log.Fatalf("Cannot resolve local listen target for Peer Probe %v: %v", peer, err)
}
peerAddress, err := net.ResolveUDPAddr("udp", peer)
if err != nil {
log.Fatalf("Cannot resolve peer target for Peer %v: %v", peer, err)
}
sendingSocket, err := net.ListenUDP("udp", sendingAddr)
if err != nil {
log.Fatalf("Cannot bind to listen for probe replies for Peer %v: %v", peer, err)
}
// Find the interface name of the sendingIP
interfaceName := findInterfaceNameFromIP(sendingIP)
rawFile, _ := sendingSocket.File()
err = syscall.BindToDevice(int(rawFile.Fd()), interfaceName)
if err != nil {
log.Printf("Unable to call BindToDevice, the replies may not be sent from the right interface!")
}
replies := make(chan probePacket, 10)
err = sockettimestamp.EnableRXTimestampsOnSocket(sendingSocket)
timestampingEnabled := false
if err != nil {
log.Printf("Unable to enable SO_TIMESTAMP on socket: %v", err)
} else {
timestampingEnabled = true
}
lossTrackingRing := PacketRingBuffer{}
lagTrackingRing := LatencyRingBuffer{}
// We want to track around 10 seconds of loss on avg, so
// we will need to store 10 seconds worth of IDs in the
// ring buffer, hence 10 s / 100 ms = 100 slots (by default)
lossTrackingRing.Setup(int(lossTrackWindow / pollRate))
lagTrackingRing.Setup(int(lossTrackWindow / pollRate))
// Start RX loop that only accepts packets from the actual target
// we are probing for.
go func() {
for {
networkBuffer := bufferPool.Get()
// networkBuffer := make([]byte, 1500)
// oob := make([]byte, 1000)
oob := bufferPool.Get()
netRead, oobN, _, rxHost, err := sendingSocket.ReadMsgUDP(networkBuffer, oob)
if err != nil {
return // TODO; handle correctly
}
var now time.Time
if timestampingEnabled {
t, err := sockettimestamp.DecodeRXTimestampFromOOB(oob[:oobN])
if err == nil {
now = t
} else {
log.Printf("Failed to DecodeRXTimestampFromOOB: %v", err)
now = time.Now()
}
} else {
now = time.Now()
}
bufferPool.Put(oob)
if !rxHost.IP.Equal(peerAddress.IP) {
metricBadPackets.WithLabelValues(`probe-reply-wrong-ip`).Inc()
continue
}
if rxHost.Port != peerAddress.Port {
metricBadPackets.WithLabelValues(`probe-reply-wrong-port`).Inc()
continue
}
replies <- unMakeProbePacket(peerProbeReply{
buf: networkBuffer[:netRead],
arrived: now,
})
bufferPool.Put(networkBuffer)
}
}()
packetN := 1 // a 64bit unit that tracks how many packets this prober has sent out
for {
select {
case PP := <-replies:
lossTrackingRing.Write(PP.Seq)
lagTrackingRing.Write(uint64(PP.Latency.Microseconds()))
case <-poll.C:
packet := makeProbePacket(uint64(packetN))
n, _, err := sendingSocket.WriteMsgUDP(
packet[:1024],
nil,
peerAddress,
)
if n != 1024 {
log.Printf("Failed to send full probe packet! %d (sent), %d (wanted) / %v", n, len(packet), err)
continue
}
if err != nil {
log.Fatalf("Failed to send probe packet! %v, aborting", err)
}
packetN++
bufferPool.Put(packet)
case <-updateMetricsTick.C:
loss := lossTrackingRing.GetPacketLoss(uint64(packetN))
metricFlowLoss.WithLabelValues(
runtimeConfig.ResolveFriendlyName(sendingAddr.IP),
runtimeConfig.ResolveFriendlyName(peerAddress.IP),
fmt.Sprint(peerAddress.Port)).Set(loss)
lag := lagTrackingRing.GetAvgLatency()
metricFlowLatency.WithLabelValues(
runtimeConfig.ResolveFriendlyName(sendingAddr.IP),
runtimeConfig.ResolveFriendlyName(peerAddress.IP),
fmt.Sprint(peerAddress.Port)).Set(float64(lag))
}
}
}
func findInterfaceNameFromIP(sendingIP string) (interfaceName string) {
interfaces, err := net.Interfaces()
if err != nil {
log.Fatalf("Could not list network interfaces, cannot intelligently bind() on ports: %v", err)
}
for _, iface := range interfaces {
addrs, err := iface.Addrs()
if err != nil {
continue
}
for _, v := range addrs {
interfaceIP, ok := v.(*net.IPNet)
if ok {
if interfaceIP.IP.String() == sendingIP {
interfaceName = iface.Name
}
}
}
}
return
}
func makeProbePacket(N uint64) []byte {
b := bufferPool.Get()
buf := &bytes.Buffer{}
// Out of the 1024 bytes in this, we can only use 512,
// as the remote side will chop off half of the packet
// when sending it back (for amp reduction design)
buf.Write([]byte(protocolMagicValue))
binary.Write(buf, binary.BigEndian, N)
ts := time.Now().UnixMicro()
binary.Write(buf, binary.BigEndian, ts)
copy(b, buf.Bytes())
return b
}
const (
// 0x8330 (The protocol magic number, also 8330 is the ASN of the entity that comissioned this project)
protocolMagicValue = "\x83\x30"
)
type probePacket struct {
Magic [2]byte
Seq uint64
Latency time.Duration
}
func unMakeProbePacket(p peerProbeReply) probePacket {
buf := bytes.NewBuffer(p.buf)
// Out of the 1024 bytes in this, we can only use 512,
// as the remote side will chop off half of the packet
// when sending it back (for amp reduction design)
PP := probePacket{}
binary.Read(buf, binary.BigEndian, &PP.Magic)
binary.Read(buf, binary.BigEndian, &PP.Seq)
ts := uint64(0)
binary.Read(buf, binary.BigEndian, &ts)
PP.Latency = p.arrived.Sub(time.UnixMicro(int64(ts)))
return PP
}
func startUDPProbeReplier(l *net.UDPConn) {
log.Printf("Listening on %v", l.LocalAddr().String())
defer log.Printf("No longer listening on %v", l.LocalAddr().String())
for {
// TODO: This (networkBuffer) could be higher with fragments
networkBuffer := bufferPool.Get()
// networkBuffer := make([]byte, 1500)
finalReadSize, sourceIP, err := l.ReadFromUDP(networkBuffer)
// captureTime := time.Now() // Capture the rx time ASAP
if err != nil {
break
}
if finalReadSize < 512 {
// To make reflection unappealing, we require clients to send us
// large payloads so we can send back smaller ones. This packet
// is simply too small, and will be dropped
metricBadPackets.WithLabelValues(`replier-too-small`).Inc()
continue
}
if !runtimeConfig.IsAllowedCIDR(sourceIP.IP) {
metricBadPackets.WithLabelValues(`replier-not-allowed-cidr`).Inc()
continue
}
truePacketSlice := networkBuffer[:finalReadSize]
inboundMagic := string(truePacketSlice[:2])
if inboundMagic != protocolMagicValue {
// This is not a XPING packet then, since it does not start with the protocol magic number
metricBadPackets.WithLabelValues(`replier-no-magic`).Inc()
continue
}
// Reply back with the packet, but at half the size to reduce the risk of us being used as UDP reflection.
_, _, err = l.WriteMsgUDP(truePacketSlice[:finalReadSize/2], nil, sourceIP)
if err != nil {
log.Printf("debug: startUDPProbeReplier/l.WriteMsgUDP: err: %v", err)
}
bufferPool.Put(networkBuffer)
}
}
func init() {
prometheus.MustRegister(metricFlowLoss)
prometheus.MustRegister(metricFlowLatency)
prometheus.MustRegister(metricBadPackets)
}
var metricFlowLoss = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Name: "xping_peer_loss_per_flow",
Help: "sliding window fraction of packetloss on {local,peer,peerport} L4 flow",
},
[]string{"local", "peer", "port"},
)
var metricFlowLatency = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Name: "xping_peer_latency_per_flow",
Help: "sliding window round-trip latency on {local,peer,peerport} L4 flow, in usec",
},
[]string{"local", "peer", "port"},
)
var metricBadPackets = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Name: "xping_bad_packets",
Help: "invalid ixp-xping packets received",
},
[]string{"reason"},
)