0% CPU usage on docker container

[ariaieboy]

Pulse Version

1.0.0-beta5

Laravel Version

10.x (latest)

PHP Version

8.2 (latest)

Livewire Version

3

Database Driver & Version

Postgresql 16 on docker container

Description

In a dockerized environment Laravel Pulse always reports 0% CPU usage.

Steps To Reproduce

create a laravel project using Laravel Sail and check the CPU usage.

[ariaieboy]

Ok, I checked the source and found that pulse uses the top command to check the CPU usage.
The top command in my containers reports 0% CPU usage.
I use Alpine Linux as the base image.

[jbrooksuk]

It looks like this may be a limitation within top, that prevents it from being able to properly read CPU usage of virtualized resources. moby/moby#26113

[ariaieboy]

It looks like this may be a limitation within top, that prevents it from being able to properly read CPU usage of virtualized resources. moby/moby#26113

I didn't test this on other base images like Ubuntu but using mpstat I can get the CPU usage in my container.

The problem with mpstat is that it shows the CPU average usage since the last reboot. and I don't know how we can use mpstat to calculate current CPU usage. maybe using mpstat 1 2 will helps but I don't know that it's a good replacement for this use case or not.

[jessarcher]

FWIW top seems to work fine in my Ubuntu container (at least on a Linux host). What host operating system are you using?

[ariaieboy]

/etc/os-release host machine

PRETTY_NAME="Ubuntu 22.04.3 LTS"
NAME="Ubuntu"
VERSION_ID="22.04"
VERSION="22.04.3 LTS (Jammy Jellyfish)"
VERSION_CODENAME=jammy
ID=ubuntu
ID_LIKE=debian
HOME_URL="https://www.ubuntu.com/"
SUPPORT_URL="https://help.ubuntu.com/"
BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
UBUNTU_CODENAME=jammy

/etc/os-release container

NAME="Alpine Linux"
ID=alpine
VERSION_ID=3.18.5
PRETTY_NAME="Alpine Linux v3.18"
HOME_URL="https://alpinelinux.org/"
BUG_REPORT_URL="https://gitlab.alpinelinux.org/alpine/aports/-/issues"

[ariaieboy]

@jessarcher I don't know how but the top 0% CPU usage is fixed now here is the output of my top command:

/var/www/html # top -bn1
Mem: 9138956K used, 1041152K free, 808552K shrd, 419480K buff, 4727092K cached
CPU:  12% usr   2% sys   0% nic  82% idle   0% io   0% irq   0% sirq
Load average: 0.56 0.50 0.47 1/813 116610
  PID  PPID USER     STAT   VSZ %VSZ CPU %CPU COMMAND
116593   44 root     S     841m   8%   1   5% {php} swoole_http_server: worker process for x
116560   44 root     S     842m   8%   0   0% {php} swoole_http_server: worker process for x
115219   44 root     S     841m   8%   0   0% {php} swoole_http_server: task worker process for x
115941   44 root     S     841m   8%   1   0% {php} swoole_http_server: task worker process for x
115684   44 root     S     841m   8%   0   0% {php} swoole_http_server: task worker process for x
115768   44 root     S     841m   8%   0   0% {php} swoole_http_server: task worker process for x
   42    27 root     S     794m   8%   1   0% {php} swoole_http_server: master process for x
   44    42 root     S     788m   8%   0   0% {php} swoole_http_server: manager process for x
   66    53 root     S     781m   8%   0   0% /usr/local/bin/php artisan horizon:work redis --name=default --supervisor=334f0c990a46-qlZY:supervisor-1 --backoff=0 --max-time=0 --max-jobs=0 --memory=128 --queue=media --sleep=3 --timeout=120 --tries=3 --rest=0 --force
116415   53 root     S     779m   8%   2   0% /usr/local/bin/php artisan horizon:work redis --name=default --supervisor=334f0c990a46-qlZY:supervisor-1 --backoff=0 --max-time=0 --max-jobs=0 --memory=128 --queue=default --sleep=3 --timeout=120 --tries=3 --rest=0 --force
   28    25 root     S     779m   8%   1   0% /usr/bin/php /var/www/html/artisan pulse:check
115435   53 root     S     779m   8%   2   0% /usr/local/bin/php artisan horizon:work redis --name=default --supervisor=334f0c990a46-qlZY:supervisor-1 --backoff=0 --max-time=0 --max-jobs=0 --memory=128 --queue=sms --sleep=3 --timeout=120 --tries=3 --rest=0 --force
   53    26 root     S     779m   8%   2   0% /usr/local/bin/php artisan horizon:supervisor 334f0c990a46-qlZY:supervisor-1 redis --workers-name=default --balance=auto --max-processes=4 --min-processes=1 --nice=0 --balance-cooldown=3 --balance-max-shift=1 --parent-id=26 --auto-scaling-strategy=time --backoff=0 --max-time=0 --max-jobs=0 --memory=128 --queue=sms,default,media --sleep=3 --timeout=120 --tries=3 --rest=0 --force
   26    25 root     S     779m   8%   3   0% /usr/bin/php /var/www/html/artisan horizon
   27    25 root     S     779m   8%   1   0% /usr/bin/php -d variables_order=EGPCS /var/www/html/artisan octane:start --server=swoole --host=0.0.0.0 --port=8000 --max-requests=250 --workers=2
   25     1 root     S    29448   0%   1   0% {supervisord} /usr/bin/python3 /usr/bin/supervisord -c /etc/supervisor/conf.d/supervisord.conf
104359    0 root     S     1704   0%   1   0% sh
115562    0 root     S     1704   0%   1   0% sh
112068    0 root     S     1704   0%   2   0% sh
111448    0 root     S     1704   0%   3   0% sh
116437    0 root     S     1676   0%   3   0% sh
   29    25 root     S     1608   0%   0   0% /bin/sh -c while [ true ]; do (php /var/www/html/artisan schedule:run --verbose --no-interaction &); sleep 60; done
    1     0 root     S     1604   0%   3   0% {start-container} /bin/sh /usr/local/bin/start-container
116610 116437 root   R     1600   0%   2   0% top -bn1
116500   29 root     S     1588   0%   2   0% sleep 60

still the laravel pulse reports 0% CPU usage.

[jessarcher]

Interesting - that output is formatted quite differently from what we're expecting. Most notably, the CPU line begins with "CPU:" but Pulse is expecting it to begin with "%Cpu(s):".

Here's the output from top on my Linux laptop:

➜  ~ top -bn1
top - 08:48:45 up 3 days, 15:20,  2 users,  load average: 1.69, 1.36, 1.22
Tasks: 579 total,   1 running, 573 sleeping,   0 stopped,   5 zombie
%Cpu(s):  9.1 us,  4.5 sy,  0.0 ni, 86.4 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
MiB Mem :  31737.7 total,   1337.3 free,  17214.1 used,  15534.1 buff/cache
MiB Swap:   8192.0 total,   7779.7 free,    412.2 used.  14523.6 avail Mem

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND

What version of top do you have?

On my host machine I have:

➜  ~ top -V
top from procps-ng 4.0.3

And in my Sail (Ubuntu) container I have:

sail@57830d912d37:/var/www/html$ top -v
  procps-ng 3.3.17

Interestingly, they both have a different CLI option for getting the version, but both follow the same format for the CPU line.

I'm guessing your version of top probably isn't the procps-ng version but rather an entirely different implementation that Alpine has chosen. The output looks more similar to the implementation Darwin ships with, although interestingly yours still supports the -bn1 flags.

[ariaieboy]

@jessarcher yes the alpine uses BusyBox 1.36.1 for top command.

[jessarcher]

Hi @ariaieboy, I've created #236, which updates the parsing of the top output to hopefully work for both versions of top on Linux.

[ariaieboy]

@jessarcher Can we have this in a tagged release so I can test it on my instances?

[timacdonald]

v1.0.0-beta8 has just been tagged.