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

NAME

6       drsnoop - Trace direct reclaim events. Uses Linux eBPF/bcc.
7

SYNOPSIS

9       drsnoop.py  [-h] [-T] [-U] [-p PID] [-t TID] [-u UID] [-d DURATION] [-n
10       name] [-v]
11

DESCRIPTION

13       drsnoop trace direct reclaim events, showing which processes are alloc‐
14       ing  pages  with  direct reclaiming. This can be useful for discovering
15       when allocstall (/p- roc/vmstat) continues to increase, whether  it  is
16       caused by some critical proc- esses or not.
17
18       This  works  by  tracing  the direct reclaim events using kernel trace‐
19       points.
20
21       This makes use of a Linux 4.4  feature  (bpf_perf_event_output());  for
22       kernels  older than 4.4, see the version under tools/old, which uses an
23       older mechanism.
24
25       Since this uses BPF, only the root user can use this tool.
26

REQUIREMENTS

28       CONFIG_BPF and bcc.
29

OPTIONS

31       -h     Print usage message.
32
33       -T     Include a timestamp column.
34
35       -U     Show UID.
36
37       -p PID Trace this process ID only (filtered in-kernel).
38
39       -t TID Trace this thread ID only (filtered in-kernel).
40
41       -u UID Trace this UID only (filtered in-kernel).
42
43       -d DURATION
44              Total duration of trace in seconds.
45
46       -n name
47              Only print processes where its name partially matches 'name'  -v
48              verbose Run in verbose mode. Will output system memory state
49
50       -v     show system memory state
51

EXAMPLES

53       Trace all direct reclaim events:
54              # drsnoop
55
56       Trace all direct reclaim events, for 10 seconds only:
57              # drsnoop -d 10
58
59       Trace all direct reclaim events, and include timestamps:
60              # drsnoop -T
61
62       Show UID:
63              # drsnoop -U
64
65       Trace PID 181 only:
66              # drsnoop -p 181
67
68       Trace UID 1000 only:
69              # drsnoop -u 1000
70
71       Trace all direct reclaim events from processes where its name partially
72       match-
73              es 'mond': # drnsnoop -n mond
74

FIELDS

76       TIME(s)
77              Time of the call, in seconds.
78
79       UID    User ID
80
81       PID    Process ID
82
83       TID    Thread ID
84
85       COMM   Process name
86

OVERHEAD

88       This traces the kernel direct reclaim tracepoints and prints output for
89       each  event.  As  the  rate  of this is generally expected to be low (<
90       1000/s), the overhead is also expected to be negligible.
91

SOURCE

93       This is from bcc.
94
95              https://github.com/iovisor/bcc
96
97       Also look in the bcc distribution for a  companion  _examples.txt  file
98       containing example usage, output, and commentary for this tool.
99

OS

101       Linux
102

STABILITY

104       Unstable - in development.
105

AUTHOR

107       Wenbo Zhang
108
109
110
111USER COMMANDS                     2019-02-20                        drsnoop(8)