zhenwei pi | e42ac41 | 2020-08-23 19:17:52 +0800 | [diff] [blame] | 1 | Demonstrations of tcprtt, the Linux eBPF/bcc version. |
| 2 | |
| 3 | |
| 4 | This program traces TCP RTT(round-trip time) to analyze the quality of |
| 5 | network, then help us to distinguish the network latency trouble is from |
| 6 | user process or physical network. |
| 7 | |
| 8 | For example, wrk show the http request latency distribution: |
| 9 | # wrk -d 30 -c 10 --latency http://192.168.122.100/index.html |
| 10 | Running 30s test @ http://192.168.122.100/index.html |
| 11 | 2 threads and 10 connections |
| 12 | Thread Stats Avg Stdev Max +/- Stdev |
| 13 | Latency 86.75ms 153.76ms 1.54s 90.85% |
| 14 | Req/Sec 160.91 76.07 424.00 67.06% |
| 15 | Latency Distribution |
| 16 | 50% 14.55ms |
| 17 | 75% 119.21ms |
| 18 | 90% 230.22ms |
| 19 | 99% 726.90ms |
| 20 | 9523 requests in 30.02s, 69.62MB read |
| 21 | Socket errors: connect 0, read 0, write 0, timeout 1 |
| 22 | |
| 23 | During wrk testing, run tcprtt: |
| 24 | # ./tcprtt -i 1 -d 10 -m |
| 25 | Tracing TCP RTT... Hit Ctrl-C to end. |
| 26 | msecs : count distribution |
| 27 | 0 -> 1 : 4 | | |
| 28 | 2 -> 3 : 0 | | |
| 29 | 4 -> 7 : 1055 |****************************************| |
| 30 | 8 -> 15 : 26 | | |
| 31 | 16 -> 31 : 0 | | |
| 32 | 32 -> 63 : 0 | | |
| 33 | 64 -> 127 : 18 | | |
| 34 | 128 -> 255 : 14 | | |
| 35 | 256 -> 511 : 14 | | |
| 36 | 512 -> 1023 : 12 | | |
| 37 | |
| 38 | The wrk output shows that the latency of web service is not stable, and tcprtt |
| 39 | also shows unstable TCP RTT. So in this situation, we need to make sure the |
| 40 | quality of network is good or not firstly. |
| 41 | |
| 42 | |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 43 | Use filter for address and(or) port. Ex, only collect local address 192.168.122.200 |
| 44 | and remote address 192.168.122.100 and remote port 80. |
zhenwei pi | e42ac41 | 2020-08-23 19:17:52 +0800 | [diff] [blame] | 45 | # ./tcprtt -i 1 -d 10 -m -a 192.168.122.200 -A 192.168.122.100 -P 80 |
| 46 | |
| 47 | |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 48 | Tracing at server side, show each clients with its own histogram. |
| 49 | For example, run tcprtt on a storage node to show initiators' rtt histogram: |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 50 | # ./tcprtt -i 1 --lport 3260 --byraddr -e |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 51 | Tracing TCP RTT... Hit Ctrl-C to end. |
| 52 | |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 53 | |
| 54 | Remote Addres = 10.194.87.206 [AVG 170] |
| 55 | usecs : count distribution |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 56 | 0 -> 1 : 0 | | |
| 57 | 2 -> 3 : 0 | | |
| 58 | 4 -> 7 : 0 | | |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 59 | 8 -> 15 : 0 | | |
| 60 | 16 -> 31 : 0 | | |
| 61 | 32 -> 63 : 31 | | |
| 62 | 64 -> 127 : 5150 |******************* | |
| 63 | 128 -> 255 : 10327 |****************************************| |
| 64 | 256 -> 511 : 1014 |*** | |
| 65 | 512 -> 1023 : 10 | | |
| 66 | 1024 -> 2047 : 7 | | |
| 67 | 2048 -> 4095 : 14 | | |
| 68 | 4096 -> 8191 : 10 | | |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 69 | |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 70 | Remote Addres = 10.194.87.197 [AVG 4293] |
| 71 | usecs : count distribution |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 72 | 0 -> 1 : 0 | | |
| 73 | 2 -> 3 : 0 | | |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 74 | 4 -> 7 : 0 | | |
| 75 | 8 -> 15 : 0 | | |
| 76 | 16 -> 31 : 0 | | |
| 77 | 32 -> 63 : 0 | | |
| 78 | 64 -> 127 : 0 | | |
| 79 | 128 -> 255 : 0 | | |
| 80 | 256 -> 511 : 0 | | |
| 81 | 512 -> 1023 : 0 | | |
| 82 | 1024 -> 2047 : 3 |******** | |
| 83 | 2048 -> 4095 : 12 |********************************** | |
| 84 | 4096 -> 8191 : 14 |****************************************| |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 85 | |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 86 | Remote Addres = 10.194.88.148 [AVG 6215] |
| 87 | usecs : count distribution |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 88 | 0 -> 1 : 0 | | |
| 89 | 2 -> 3 : 0 | | |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 90 | 4 -> 7 : 0 | | |
| 91 | 8 -> 15 : 0 | | |
| 92 | 16 -> 31 : 0 | | |
| 93 | 32 -> 63 : 0 | | |
| 94 | 64 -> 127 : 0 | | |
| 95 | 128 -> 255 : 0 | | |
| 96 | 256 -> 511 : 0 | | |
| 97 | 512 -> 1023 : 0 | | |
| 98 | 1024 -> 2047 : 0 | | |
| 99 | 2048 -> 4095 : 0 | | |
| 100 | 4096 -> 8191 : 2 |****************************************| |
| 101 | |
| 102 | Remote Addres = 10.194.87.90 [AVG 2188] |
| 103 | usecs : count distribution |
| 104 | 0 -> 1 : 0 | | |
| 105 | 2 -> 3 : 0 | | |
| 106 | 4 -> 7 : 0 | | |
| 107 | 8 -> 15 : 0 | | |
| 108 | 16 -> 31 : 0 | | |
| 109 | 32 -> 63 : 0 | | |
| 110 | 64 -> 127 : 0 | | |
| 111 | 128 -> 255 : 0 | | |
| 112 | 256 -> 511 : 15 |********* | |
| 113 | 512 -> 1023 : 30 |****************** | |
| 114 | 1024 -> 2047 : 50 |****************************** | |
| 115 | 2048 -> 4095 : 65 |****************************************| |
| 116 | 4096 -> 8191 : 22 |************* | |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 117 | |
| 118 | .... |
| 119 | |
| 120 | |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 121 | Use -e(--extension) to show extension RTT: |
| 122 | # ./tcprtt -i 1 -e |
| 123 | |
| 124 | All Addresses = ******* [AVG 324] |
| 125 | usecs : count distribution |
| 126 | 0 -> 1 : 0 | | |
| 127 | 2 -> 3 : 0 | | |
| 128 | 4 -> 7 : 0 | | |
| 129 | 8 -> 15 : 0 | | |
| 130 | 16 -> 31 : 0 | | |
| 131 | 32 -> 63 : 0 | | |
| 132 | 64 -> 127 : 5360 |******** | |
| 133 | 128 -> 255 : 23834 |****************************************| |
| 134 | 256 -> 511 : 11276 |****************** | |
| 135 | 512 -> 1023 : 700 |* | |
| 136 | 1024 -> 2047 : 434 | | |
| 137 | 2048 -> 4095 : 356 | | |
| 138 | 4096 -> 8191 : 328 | | |
| 139 | 8192 -> 16383 : 91 | | |
| 140 | |
| 141 | |
zhenwei pi | e42ac41 | 2020-08-23 19:17:52 +0800 | [diff] [blame] | 142 | Full USAGE: |
| 143 | |
| 144 | # ./tcprtt -h |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 145 | usage: tcprtt [-h] [-i INTERVAL] [-d DURATION] [-T] [-m] [-p LPORT] |
| 146 | [-P RPORT] [-a LADDR] [-A RADDR] [-b] [-B] [-e] [-D] |
Hariharan Ananthakrishnan | 04893e3 | 2021-08-12 05:55:21 -0700 | [diff] [blame] | 147 | [-4 | -6] |
zhenwei pi | e42ac41 | 2020-08-23 19:17:52 +0800 | [diff] [blame] | 148 | |
| 149 | Summarize TCP RTT as a histogram |
| 150 | |
| 151 | optional arguments: |
| 152 | -h, --help show this help message and exit |
| 153 | -i INTERVAL, --interval INTERVAL |
| 154 | summary interval, seconds |
| 155 | -d DURATION, --duration DURATION |
| 156 | total duration of trace, seconds |
| 157 | -T, --timestamp include timestamp on output |
| 158 | -m, --milliseconds millisecond histogram |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 159 | -p LPORT, --lport LPORT |
| 160 | filter for local port |
| 161 | -P RPORT, --rport RPORT |
| 162 | filter for remote port |
| 163 | -a LADDR, --laddr LADDR |
| 164 | filter for local address |
| 165 | -A RADDR, --raddr RADDR |
| 166 | filter for remote address |
| 167 | -b, --byladdr show sockets histogram by local address |
| 168 | -B, --byraddr show sockets histogram by remote address |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 169 | -e, --extension show extension summary(average) |
zhenwei pi | e42ac41 | 2020-08-23 19:17:52 +0800 | [diff] [blame] | 170 | -D, --debug print BPF program before starting (for debugging |
| 171 | purposes) |
Hariharan Ananthakrishnan | 04893e3 | 2021-08-12 05:55:21 -0700 | [diff] [blame] | 172 | -4, --ipv4 trace IPv4 family only |
| 173 | -6, --ipv6 trace IPv6 family only |
zhenwei pi | e42ac41 | 2020-08-23 19:17:52 +0800 | [diff] [blame] | 174 | |
| 175 | examples: |
| 176 | ./tcprtt # summarize TCP RTT |
| 177 | ./tcprtt -i 1 -d 10 # print 1 second summaries, 10 times |
| 178 | ./tcprtt -m -T # summarize in millisecond, and timestamps |
zhenwei pi | 4cbcd9a | 2020-09-29 00:07:38 +0800 | [diff] [blame] | 179 | ./tcprtt -p # filter for local port |
| 180 | ./tcprtt -P # filter for remote port |
| 181 | ./tcprtt -a # filter for local address |
| 182 | ./tcprtt -A # filter for remote address |
| 183 | ./tcprtt -b # show sockets histogram by local address |
| 184 | ./tcprtt -B # show sockets histogram by remote address |
zhenwei pi | e42ac41 | 2020-08-23 19:17:52 +0800 | [diff] [blame] | 185 | ./tcprtt -D # show debug bpf text |
zhenwei pi | 5057944 | 2021-06-21 15:06:58 +0800 | [diff] [blame] | 186 | ./tcprtt -e # show extension summary(average) |
Hariharan Ananthakrishnan | 04893e3 | 2021-08-12 05:55:21 -0700 | [diff] [blame] | 187 | ./tcprtt -4 # trace IPv4 family only |
| 188 | ./tcprtt -6 # trace IPv6 family only |