Sep 10, 2016

Monitoring Linux Performance with BCC and BPF Dynamic Tracing Tools


This article will guide you through the steps to monitor Linux performance, networking and much more using BCC (BPF Compiler Collection) and BPF (Berkeley Packet Filtersdynamic tracing tools.






Installing BCC on Linux Machines
To begin with the installation, first you need to compile linux kernel with the flags set as below:
CONFIG_BPF=y
CONFIG_BPF_SYSCALL=y
# [optional, for tc filters]
CONFIG_NET_CLS_BPF=m
# [optional, for tc actions]
CONFIG_NET_ACT_BPF=m
CONFIG_BPF_JIT=y
CONFIG_HAVE_BPF_JIT=y
# [optional, for kprobes]
CONFIG_BPF_EVENTS=y
To verify your linux kernel flags, view the file /proc/config.gz or execute the following commands:
grep CONFIG_BPF= /boot/config-`uname -r`
CONFIG_BPF=y
grep CONFIG_BPF_SYSCALL= /boot/config-`uname -r`
CONFIG_BPF_SYSCALL=y
grep CONFIG_NET_CLS_BPF= /boot/config-`uname -r`
CONFIG_NET_CLS_BPF=m
grep CONFIG_NET_ACT_BPF= /boot/config-`uname -r`
CONFIG_NET_ACT_BPF=m
grep CONFIG_BPF_JIT= /boot/config-`uname -r`
CONFIG_BPF_JIT=y
grep CONFIG_HAVE_BPF_JIT= /boot/config-`uname -r`
CONFIG_HAVE_BPF_JIT=y
grep CONFIG_BPF_EVENTS= /boot/config-`uname -r`
CONFIG_BPF_EVENTS=y
Once done verifying kernel flags, start installing BCC tools on your Linux machines.

Installing BCC on Ubuntu 16.04
Installation steps are very straightforward. You don't need to upgrade kernel or compile it from source.
$ echo "deb [trusted=yes] https://repo.iovisor.org/apt/xenial xenial-nightly main" | sudo tee /etc/apt/sources.list.d/iovisor.list
$ sudo apt-get update
$ sudo apt-get install bcc-tools
Installing BCC on Ubuntu 14.04
First you need to install 4.3+ Linux kernel, from http://kernel.ubuntu.com/~kernel-ppa/mainline.
You can create a small shell script “bcc-install-ubuntu.sh” with the following contents.
Note: update PREFIX value to the latest date, and also browse the files in the PREFIX url provided to get the actual REL value, substitute them in the shell script.
$ vi bcc-install-ubuntu.sh
#!/bin/bash
VER=4.5.1-040501
PREFIX=http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.5.1-wily/
REL=201604121331
wget ${PREFIX}/linux-headers-${VER}-generic_${VER}.${REL}_amd64.deb
wget ${PREFIX}/linux-headers-${VER}_${VER}.${REL}_all.deb
wget ${PREFIX}/linux-image-${VER}-generic_${VER}.${REL}_amd64.deb
sudo dpkg -i linux-*${VER}.${REL}*.deb
Save and exit. 
Now make it executable by executing the following command:
$ chmod +x bcc-install-ubuntu.sh
$ sh bcc-install-ubuntu.sh
reboot your system.
$ reboot
Once reboot completed, run the following commands to install signed BCC packages:
$ sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys D4284CDD
$ echo "deb https://repo.iovisor.org/apt trusty main" | sudo tee /etc/apt/sources.list.d/iovisor.list
$ sudo apt-get update
$ sudo apt-get install binutils bcc bcc-tools libbcc-examples python-bcc
Installing BCC On Fedora 24-23
First you need to install 4.2+ kernel from http://alt.fedoraproject.org/pub/alt/rawhide-kernel-nodebug, if your machine running a version lower than what is required. 
$ sudo dnf config-manager --add-repo=http://alt.fedoraproject.org/pub/alt/rawhide-kernel-nodebug/fedora-rawhide-kernel-nodebug.repo
$ sudo dnf update
$ reboot
Add the BCC tools repository, update your system and install the tools by running the following commands:
$ echo -e '[iovisor]\nbaseurl=https://repo.iovisor.org/yum/nightly/f23/$basearch\nenabled=1\ngpgcheck=0' | sudo tee /etc/yum.repos.d/iovisor.repo
$ sudo dnf update
$ sudo dnf install bcc-tools
Installing BCC On Arch Linux – AUR
You need to upgrade its kernel to at least version 4.3.1-1, and then install the packages below using any Arch package managers such as pacaur, yaourt, cower, etc.
bcc bcc-tools python-bcc python2-bcc

Using BCC Tools in Linux
All the BCC tools are installed under /usr/share/bcc/tools directory. However, you can alternatively run them from the BCC Github repository under /tools where they end with a .py extension. We will show you a few examples under – monitoring general Linux system's performance and networking.

$ ls /usr/share/bcc/tools 
argdist       capable     filetop         offwaketime  stackcount  vfscount
bashreadline  cpudist     funccount       old          stacksnoop  vfsstat
biolatency    dcsnoop     funclatency     oomkill      statsnoop   wakeuptime
biosnoop      dcstat      gethostlatency  opensnoop    syncsnoop   xfsdist
biotop        doc         hardirqs        pidpersec    tcpaccept   xfsslower
bitesize      execsnoop   killsnoop       profile      tcpconnect  zfsdist
btrfsdist     ext4dist    mdflush         runqlat      tcpconnlat  zfsslower
btrfsslower   ext4slower  memleak         softirqs     tcpretrans
cachestat     filelife    mysqld_qslower  solisten     tplist
cachetop      fileslower  offcputime      sslsniff     trace
Trace open() syscalls
Let’s begin by tracing all open() syscalls using opensnoop and see how various applications functions by identifying their data files, config files and much more:
$ cd /usr/share/bcc/tools 
$ sudo ./opensnoop
PID    COMM               FD ERR PATH
1      systemd            35   0 /proc/self/mountinfo
2797   udisksd            13   0 /proc/self/mountinfo
1      systemd            35   0 /sys/devices/pci0000:00/0000:00:0d.0/ata3/host2/target2:0:0/2:0:0:0/block/sda/sda1/uevent
1      systemd            35   0 /run/udev/data/b8:1
1      systemd            -1   2 /etc/systemd/system/sys-kernel-debug-tracing.mount
1      systemd            -1   2 /run/systemd/system/sys-kernel-debug-tracing.mount
1      systemd            -1   2 /run/systemd/generator/sys-kernel-debug-tracing.mount
1      systemd            -1   2 /usr/local/lib/systemd/system/sys-kernel-debug-tracing.mount
2247   systemd            15   0 /proc/self/mountinfo
1      systemd            -1   2 /lib/systemd/system/sys-kernel-debug-tracing.mount
1      systemd            -1   2 /usr/lib/systemd/system/sys-kernel-debug-tracing.mount
1      systemd            -1   2 /run/systemd/generator.late/sys-kernel-debug-tracing.mount
1      systemd            -1   2 /etc/systemd/system/sys-kernel-debug-tracing.mount.wants
1      systemd            -1   2 /etc/systemd/system/sys-kernel-debug-tracing.mount.requires
1      systemd            -1   2 /run/systemd/system/sys-kernel-debug-tracing.mount.wants
1      systemd            -1   2 /run/systemd/system/sys-kernel-debug-tracing.mount.requires
1      systemd            -1   2 /run/systemd/generator/sys-kernel-debug-tracing.mount.wants
1      systemd            -1   2 /run/systemd/generator/sys-kernel-debug-tracing.mount.requires
1      systemd            -1   2 /usr/local/lib/systemd/system/sys-kernel-debug-tracing.mount.wants
1      systemd            -1   2 /usr/local/lib/systemd/system/sys-kernel-debug-tracing.mount.requires
1      systemd            -1   2 /lib/systemd/system/sys-kernel-debug-tracing.mount.wants
1      systemd            -1   2 /lib/systemd/system/sys-kernel-debug-tracing.mount.requires
1      systemd            -1   2 /usr/lib/systemd/system/sys-kernel-debug-tracing.mount.wants
1      systemd            -1   2 /usr/lib/systemd/system/sys-kernel-debug-tracing.mount.requires
1      systemd            -1   2 /run/systemd/generator.late/sys-kernel-debug-tracing.mount.wants
1      systemd            -1   2 /run/systemd/generator.late/sys-kernel-debug-tracing.mount.requires
1      systemd            -1   2 /etc/systemd/system/sys-kernel-debug-tracing.mount.d
1      systemd            -1   2 /run/systemd/system/sys-kernel-debug-tracing.mount.d
1      systemd            -1   2 /run/systemd/generator/sys-kernel-debug-tracing.mount.d
....
Summarize Block Device I/O Latency
Following example shows a summarized distribution of disk I/O latency using bio-latency. After executing the command, wait for a few minutes and press Ctrl-C to end it and view the output.
$ sudo ./biolatecncy
Tracing block device I/O... Hit Ctrl-C to end.
^C
usecs               : count     distribution
0 -> 1          : 0        |                                        |
2 -> 3          : 0        |                                        |
4 -> 7          : 0        |                                        |
8 -> 15         : 0        |                                        |
16 -> 31         : 0        |                                        |
32 -> 63         : 0        |                                        |
64 -> 127        : 0        |                                        |
128 -> 255        : 3        |****************************************|
256 -> 511        : 3        |****************************************|
512 -> 1023       : 1        |*************                           |
Trace New Processes via exec() Syscalls
Now we will be tracing new processes in execution using execsnoop tool. Each time a process is forked by fork() and exec() syscalls, it is shown in the output. However, not all processes are captured.
$ sudo ./execsnoop
PCOMM            PID    PPID   RET ARGS
gnome-screensho  14882  14881    0 /usr/bin/gnome-screenshot --gapplication-service
systemd-hostnam  14892  1        0 /lib/systemd/systemd-hostnamed
nautilus         14897  2767    -2 /home/tecmint/bin/net usershare info
nautilus         14897  2767    -2 /home/tecmint/.local/bin/net usershare info
nautilus         14897  2767    -2 /usr/local/sbin/net usershare info
nautilus         14897  2767    -2 /usr/local/bin/net usershare info
nautilus         14897  2767    -2 /usr/sbin/net usershare info
nautilus         14897  2767    -2 /usr/bin/net usershare info
nautilus         14897  2767    -2 /sbin/net usershare info
nautilus         14897  2767    -2 /bin/net usershare info
nautilus         14897  2767    -2 /usr/games/net usershare info
nautilus         14897  2767    -2 /usr/local/games/net usershare info
nautilus         14897  2767    -2 /snap/bin/net usershare info
compiz           14899  14898   -2 /home/tecmint/bin/libreoffice --calc
compiz           14899  14898   -2 /home/tecmint/.local/bin/libreoffice --calc
compiz           14899  14898   -2 /usr/local/sbin/libreoffice --calc
compiz           14899  14898   -2 /usr/local/bin/libreoffice --calc
compiz           14899  14898   -2 /usr/sbin/libreoffice --calc
libreoffice      14899  2252     0 /usr/bin/libreoffice --calc
dirname          14902  14899    0 /usr/bin/dirname /usr/bin/libreoffice
basename         14903  14899    0 /usr/bin/basename /usr/bin/libreoffice
...

Trace Slow ext4 Operations
You can accomplish this by using ext4slower to trace the ext4 file system common operations that are slower than 10ms, which help us identify independently slow disk I/O via the file system. It only outputs those operations that exceed a threshold:
$ sudo ./execslower
Tracing ext4 operations slower than 10 ms
TIME     COMM           PID    T BYTES   OFF_KB   LAT(ms) FILENAME
11:59:13 upstart        2252   W 48      1          10.76 dbus.log
11:59:13 gnome-screensh 14993  R 144     0          10.96 settings.ini
11:59:13 gnome-screensh 14993  R 28      0          16.02 gtk.css
11:59:13 gnome-screensh 14993  R 3389    0          18.32 gtk-main.css
11:59:25 rs:main Q:Reg  1826   W 156     60         31.85 syslog
11:59:25 pool           15002  R 208     0          14.98 .xsession-errors
11:59:25 pool           15002  R 644     0          12.28 .ICEauthority
11:59:25 pool           15002  R 220     0          13.38 .bash_logout
11:59:27 dconf-service  2599   S 0       0          22.75 user.BHDKOY
11:59:33 compiz         2548   R 4096    0          19.03 firefox.desktop
11:59:34 compiz         15008  R 128     0          27.52 firefox.sh
11:59:34 firefox        15008  R 128     0          36.48 firefox
11:59:34 zeitgeist-daem 2988   S 0       0          62.23 activity.sqlite-wal
11:59:34 zeitgeist-fts  2996   R 8192    40         15.67 postlist.DB
11:59:34 firefox        15008  R 140     0          18.05 dependentlibs.list
11:59:34 zeitgeist-fts  2996   S 0       0          25.96 position.tmp
11:59:34 firefox        15008  R 4096    0          10.67 libplc4.so
11:59:34 zeitgeist-fts  2996   S 0       0          11.29 termlist.tmp
...
Trace Block Device I/O with PID and Latency
Now let's get into printing a line per disk I/O each second, with details such as process ID, sector, bytes, latency among others using biosnoop:
$ sudo ./biosnoop
TIME(s)        COMM           PID    DISK    T  SECTOR    BYTES   LAT(ms)
0.000000000    ?              0              R  -1        8          0.26
2.047897000    ?              0              R  -1        8          0.21
3.280028000    kworker/u4:0   14871  sda     W  30552896  4096       0.24
3.280271000    jbd2/sda1-8    545    sda     W  29757720  12288      0.40
3.298318000    jbd2/sda1-8    545    sda     W  29757744  4096       0.14
4.096084000    ?              0              R  -1        8          0.27
6.143977000    ?              0              R  -1        8          0.27
8.192006000    ?              0              R  -1        8          0.26
8.303938000    kworker/u4:2   15084  sda     W  12586584  4096       0.14
8.303965000    kworker/u4:2   15084  sda     W  25174736  4096       0.14
10.239961000   ?              0              R  -1        8          0.26
12.292057000   ?              0              R  -1        8          0.20
14.335990000   ?              0              R  -1        8          0.26
16.383798000   ?              0              R  -1        8          0.17
...

Trace Page Cache hit/miss Ratio

Using cachestat command to display one line of summarized statistics from the system cache every second. This enables for system tuning operations by pointing out low cache hit ratio and high rate of misses:
$ sudo ./cachestat
HITS   MISSES  DIRTIES  READ_HIT% WRITE_HIT%   BUFFERS_MB  CACHED_MB
0        0        0       0.0%       0.0%           19        544
4        4        2      25.0%      25.0%           19        544
1321       33        4      97.3%       2.3%           19        545
7476        0        2     100.0%       0.0%           19        545
6228       15        2      99.7%       0.2%           19        545
0        0        0       0.0%       0.0%           19        545
7391      253      108      95.3%       2.7%           19        545
33608     5382       28      86.1%      13.8%           19        567
25098       37       36      99.7%       0.0%           19        566
17624      239      416      96.3%       0.5%           19        520
...

Trace TCP Active Connections

Monitoring TCP connections every second using tcpconnect. Its output includes source and destination address, and port number. This tool is useful for tracing unexpected TCP connections, it help us to identify inefficiencies in application configurations or an attacker.
$ sudo ./tcpconnect
PID COMM IP SADDR DADDR DPORT 15272 Socket Threa 4 10.0.2.15 91.189.89.240 80 15272 Socket Threa 4 10.0.2.15 216.58.199.142 443 15272 Socket Threa 4 10.0.2.15 216.58.199.142 80 15272 Socket Threa 4 10.0.2.15 216.58.199.174 443 15272 Socket Threa 4 10.0.2.15 54.200.62.216 443 15272 Socket Threa 4 10.0.2.15 54.200.62.216 443 15272 Socket Threa 4 10.0.2.15 117.18.237.29 80 15272 Socket Threa 4 10.0.2.15 216.58.199.142 80 15272 Socket Threa 4 10.0.2.15 216.58.199.131 80 15272 Socket Threa 4 10.0.2.15 216.58.199.131 443 15272 Socket Threa 4 10.0.2.15 52.222.135.52 443 15272 Socket Threa 4 10.0.2.15 216.58.199.131 443 15272 Socket Threa 4 10.0.2.15 54.200.62.216 443 15272 Socket Threa 4 10.0.2.15 54.200.62.216 443 15272 Socket Threa 4 10.0.2.15 216.58.199.132 443 15272 Socket Threa 4 10.0.2.15 216.58.199.131 443 15272 Socket Threa 4 10.0.2.15 216.58.199.142 443 15272 Socket Threa 4 10.0.2.15 54.69.17.198 443 15272 Socket Threa 4 10.0.2.15 54.69.17.198 443 ...
All the tools above can also be used with various options, to enable the help page for a given tool, make use of the -h option, for example:
$ sudo ./tcpconnect -h
usage: tcpconnect [-h] [-t] [-p PID] [-P PORT]
Trace TCP connects
optional arguments:
-h, --help            show this help message and exit
-t, --timestamp       include timestamp on output
-p PID, --pid PID     trace this PID only
-P PORT, --port PORT  comma-separated list of destination ports to trace.
examples:
./tcpconnect           # trace all TCP connect()s
./tcpconnect -t        # include timestamps
./tcpconnect -p 181    # only trace PID 181
./tcpconnect -P 80     # only trace port 80
./tcpconnect -P 80,81  # only trace port 80 and 81

Trace Failed exec()s Syscalls

To trace failed exec()s syscalls, employ the -x option with opensnoop as below:
$ sudo ./opensnoop -x
PID    COMM               FD ERR PATH
15414  pool               -1   2 /home/.hidden
15415  (ostnamed)         -1   2 /sys/fs/cgroup/cpu/system.slice/systemd-hostnamed.service/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/cpu/system.slice/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/cpuacct/system.slice/systemd-hostnamed.service/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/cpuacct/system.slice/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/blkio/system.slice/systemd-hostnamed.service/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/blkio/system.slice/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/memory/system.slice/systemd-hostnamed.service/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/memory/system.slice/cgroup.procs
15415  (ostnamed)         -1   2 /sys/fs/cgroup/pids/system.slice/systemd-hostnamed.service/cgroup.procs
2548   compiz             -1   2 
15416  systemd-cgroups    -1   2 /run/systemd/container
15416  systemd-cgroups    -1   2 /sys/fs/kdbus/0-system/bus
15415  systemd-hostnam    -1   2 /run/systemd/container
15415  systemd-hostnam    -1  13 /proc/1/environ
15415  systemd-hostnam    -1   2 /sys/fs/kdbus/0-system/bus
1695   dbus-daemon        -1   2 /run/systemd/users/0
15415  systemd-hostnam    -1   2 /etc/machine-info
15414  pool               -1   2 /home/tecmint/.hidden
15414  pool               -1   2 /home/tecmint/Binary/.hidden
2599   dconf-service      -1   2 /run/user/1000/dconf/user
...




Trace Particular Process Functions

The last example below shows how to execute a custom trace operation. We are tracing a particular process using its PID. First determine the process ID:
$ pidof firefox
15437
Next, run the custom trace command. In the command below: -p specifies the process ID, do_sys_open() is a kernel function that is traced dynamically including its second argument as a string.
$ sudo ./trace -p 4095 'do_sys_open "%s", arg2'
TIME     PID    COMM         FUNC             -
12:17:14 15437  firefox      do_sys_open      /run/user/1000/dconf/user
12:17:14 15437  firefox      do_sys_open      /home/tecmint/.config/dconf/user
12:18:07 15437  firefox      do_sys_open      /run/user/1000/dconf/user
12:18:07 15437  firefox      do_sys_open      /home/tecmint/.config/dconf/user
12:18:13 15437  firefox      do_sys_open      /sys/devices/system/cpu/present
12:18:13 15437  firefox      do_sys_open      /dev/urandom
12:18:13 15437  firefox      do_sys_open      /dev/urandom
12:18:14 15437  firefox      do_sys_open      /usr/share/fonts/truetype/liberation/LiberationSans-Italic.ttf
12:18:14 15437  firefox      do_sys_open      /usr/share/fonts/truetype/liberation/LiberationSans-Italic.ttf
12:18:14 15437  firefox      do_sys_open      /usr/share/fonts/truetype/liberation/LiberationSans-Italic.ttf
12:18:14 15437  firefox      do_sys_open      /sys/devices/system/cpu/present
12:18:14 15437  firefox      do_sys_open      /dev/urandom
12:18:14 15437  firefox      do_sys_open      /dev/urandom
12:18:14 15437  firefox      do_sys_open      /dev/urandom
12:18:14 15437  firefox      do_sys_open      /dev/urandom
12:18:15 15437  firefox      do_sys_open      /sys/devices/system/cpu/present
12:18:15 15437  firefox      do_sys_open      /dev/urandom
12:18:15 15437  firefox      do_sys_open      /dev/urandom
12:18:15 15437  firefox      do_sys_open      /sys/devices/system/cpu/present
12:18:15 15437  firefox      do_sys_open      /dev/urandom
12:18:15 15437  firefox      do_sys_open      /dev/urandom
....
That's it.



Published: September 10, 2016 | Last Updated: January 27, 2017

Post a Comment

 
TECH SUPPORT © 2012-2017