There are several common ways to run programs in the background of Linux.
The following practice is based on the
Ubuntu2204 system, but theoretically other Linux distributions are also applicable.
1. Distinguish usage scenarios
If the command takes a little longer to execute and you don’t need to see the output, use
& as a background run flag or use
ctrl+z to hang the process in the background after running, combined with
fg to switch the background tasks of the current session.
If you want the command to stay running after the session ends, consider using
screen for short time tasks (e.g. scripts) and
nohup for long time tasks (e.g. proxies).
If there is a need for logging/permission control,
nohup are not enough, and it is appropriate to use
systemd and other daemon programs to create background daemons (web applications).
ping as an example, send 4 icmp packets to 184.108.40.206 and record the result.
The effect of the
& symbol is similar to
ctrl-z after running, so if you don’t want to record the results and want to switch back to view them easily at any time, consider using
ctrl+z to suspend the process to the background.
After hanging, use
jobs -l to view the hung background processes.
Use the command
fg %num to bring the background task process back up and you can see that the process is executing again.
The above methods are only valid for the current session, if the session ends, the program will end its run.
If you want the program to continue running after the session ends, consider using
screen, which some Linux distributions need to install themselves.
Use screen to create a ping session.
After exiting the session, recreate the session and use the create ping session command again.
You can see that the command is still running.
screen is more suitable for executing some scripts than
nohup, while nohup is suitable for running long-time tasks. nohup is called
no hang up, which means running the program without hanging up, ignoring the system’s
Hang Up signal.
Searching the web for ways to use
nohup, the more common ways to run it are as follows.
What is more puzzling here is the meaning of
There are three types of inputs and outputs defined in Linux, which are
- 0: standard input
- 1: standard output
- 2: Error output
The standard output
1 can be left unwritten at runtime, so executing
command > command.log is equivalent to
command 1 > command.log, and after understanding this, the meaning of
2>&1 becomes clearer, indicating that the error output
2 is redirected to the standard output variable
&1, and then combined with
> to be output to the log in the log.
Writing the full command is equivalent to the following.
Note: nohup command 2>&1 > command.log & is incorrectly written, the error output (2) is redirected to the standard output (1) and subsequently the standard output (1) is redirected to the file, so the error output is disabled.
As an example, take the
nc program to scan the port and execute it with nuhup as follows.
/tmp/nc.log file and you can see that it is executing.
nohup is still very good for performing simple background tasks, but if you have requirements for output log cutting, running users, booting, etc., you need to consider daemon type programs such as
supervisor is a lightweight process control system similar to
systemd, used to run programs for long periods of time, often used for web background programs, proxy programs, etc. On most Linux distributions you need to install it yourself, e.g. install and start the service in
supervisor supports very rich parameter settings, environment settings and daemons compared to
nohup. After installing supervisor, take the synchronization software
syncthing as an example.
/etc/supervisor/conf.d/syncthing.conf file, as follows.
Most of the parameters are very simple and easy to understand, where
- autorestart: boot self-start
- stdout_*: standard output log
- strerr_*: exception output log
- user: the user who runs the program
- environment: set the environment parameters for running
Makes the newly created service effective.
supervisorctl supports multiple operations, such as
- update: means update the list of services and re-run the new/modified services
- status: check the running status of all services
- start/stop/restart: start/stop/restart the service
systemd is the system suite that comes with most distributions to replace
SysV init, providing reliable parallelism during the boot process and centralized management of processes, daemons, services and mount points.
If you are familiar with modern Linux operating systems, you will most likely not be unfamiliar with
systemd is a relatively heavyweight system suite, and unless there is a clear need to use it, it is generally not considered as a way to run small applications in the background. supervisor is sufficient for most application scenarios.
systemd common user unit configuration directory is as follows (as distinct from the system unit configuration)
Take the synchronization software
syncthing as an example, add a configuration file
$HOME/.config/systemd/user/syncthing.service with the following content.
Start the service.
If you modify the configuration sheet, you need to run a refresh of systemd.
The following is a description of the key parts of the configuration file.