bcc-tcptop(8)

1tcptop(8)                   System Manager's Manual                  tcptop(8)
2
3
4

NAME

6       tcptop - Summarize TCP send/recv throughput by host. Top for TCP.
7

SYNOPSIS

9       tcptop [-h] [-C] [-S] [-p PID] [--cgroupmap MAPPATH]
10                 [--mntnsmap MAPPATH] [interval] [count]
11

DESCRIPTION

13       This is top for TCP sessions.
14
15       This  summarizes  TCP send/receive Kbytes by host, and prints a summary
16       that refreshes, along other system-wide metrics.
17
18       This uses dynamic tracing of kernel  TCP  send/receive  functions,  and
19       will need to be updated to match kernel changes.
20
21       The  traced  TCP functions are usually called at a lower rate than per-
22       packet functions, and therefore have lower overhead. The traced data is
23       summarized  in-kernel  using  a  BPF map to further reduce overhead. At
24       very high TCP event rates, the overhead may still  be  measurable.  See
25       the OVERHEAD section for more details.
26
27       Since this uses BPF, only the root user can use this tool.
28

REQUIREMENTS

30       CONFIG_BPF and bcc.
31

OPTIONS

33       -h     Print USAGE message.
34
35       -C     Don't clear the screen.
36
37       -S     Don't print the system summary line (load averages).
38
39       -p PID Trace this PID only.
40
41       --cgroupmap MAPPATH
42              Trace cgroups in this BPF map only (filtered in-kernel).
43
44       --mntnsmap  MAPPATH
45              Trace  mount  namespaces  in this BPF map only (filtered in-ker‐
46              nel).
47
48       interval
49              Interval between updates, seconds (default 1).
50
51       count  Number of interval summaries (default is many).
52

EXAMPLES

54       Summarize TCP throughput by active sessions, 1 second refresh:
55              # tcptop
56
57       Don't clear the screen (rolling output), and 5 second summaries:
58              # tcptop -C 5
59
60       Trace PID 181 only, and don't clear the screen:
61              # tcptop -Cp 181
62
63       Trace a set of cgroups only (see special_filtering.md from bcc  sources
64       for more details):
65              # tcptop --cgroupmap /sys/fs/bpf/test01
66

FIELDS

68       loadavg:
69              The contents of /proc/loadavg
70
71       PID    Process ID.
72
73       COMM   Process name.
74
75       LADDR  Local address (IPv4), and TCP port
76
77       RADDR  Remote address (IPv4), and TCP port
78
79       LADDR6 Source address (IPv6), and TCP port
80
81       RADDR6 Destination address (IPv6), and TCP port
82
83       RX_KB  Received Kbytes
84
85       TX_KB  Transmitted Kbytes
86

OVERHEAD

88       This  traces  all send/receives in TCP, high in the TCP/IP stack (close
89       to the application) which are usually called at a lower rate than  per-
90       packet  functions, lowering overhead. It also summarizes data in-kernel
91       to further reduce overhead. These techniques help, but there may  still
92       be  measurable overhead at high send/receive rates, eg, ~13% of one CPU
93       at 100k events/sec.  use funccount to count the kprobes in the tool  to
94       find  out this rate, as the overhead is relative to the rate. Some sam‐
95       ple production servers tested found total TCP event rates of 4k to  15k
96       per  second,  and  the  CPU overhead at these rates ranged from 0.5% to
97       2.0% of one CPU. If your send/receive rate is low (eg, <1000/sec)  then
98       the  overhead  is  expected to be negligible; Test in a lab environment
99       first.
100

SOURCE

102       This is from bcc.
103
104              https://github.com/iovisor/bcc
105
106       Also look in the bcc distribution for a  companion  _examples.txt  file
107       containing example usage, output, and commentary for this tool.
108

OS

110       Linux
111

STABILITY

113       Unstable - in development.
114

AUTHOR

116       Brendan Gregg
117

INSPIRATION

119       top(1) by William LeFebvre
120