How to suspend and bring a background process to foreground

As Tim said, type fg to bring the last process back to foreground.

If you have more than one process running in the background, do this:

$ jobs
[1]   Stopped                 vim
[2]-  Stopped                 bash
[3]+  Stopped                 vim 23

fg %3 to bring the vim 23 process back to foreground.

To suspend the process running in the background, use:

kill -STOP %job_id

The SIGSTOP signal stops (pauses) a process in essentially the same way Ctrl+Z does.

example: kill -STOP %3.

sources: How to send signals to processes in Linux and Unix and How to manage background and foreground jobs.

This should be true of any shell with job control, which (for the most part) you can take for granted unless dealing with a truly ancient shell. It's in the POSIX standard, so even dash supports job control (when run interactively or with -m).

Interactive

• Ctrl+z will suspend the currently foregrounded program
• bg will background the most recently suspended program
  (use bg %2 with the job number, which you can check with jobs)
• fg will foreground the most recently suspended program

In zsh, you can write a key binding to implicitly run fg from the prompt via another Ctrl+z:

_zsh_cli_fg() { fg; }
zle -N _zsh_cli_fg
bindkey '^Z' _zsh_cli_fg

There's probably also a clever way to implicitly run bg upon suspending, but it seems unwise; at least for me, the majority of my Ctrl+z usage is because Ctrl+c is failing to break out; I want to follow that with e.g. kill %1 rather than bg, and I certainly don't want to default to killing! (This logic also extends into why I no longer use this key binding: If I'm hammering on Ctrl+z to stop a process, the last thing I want it to do is resume!)

Non-interactive

If you're in a different shell instance (or a different user, sometimes including sudo commands), you likely won't be able to use job numbers.

You can still act on another process once you know its process ID (PID). You can get the PID with pgrep …, or ps aux |grep … (or from the same shell, jobs -l, or $!) and you can then run:

kill -STOP $PID  # suspend
kill -CONT $PID  # continue (resume)

If you do not know the process ID and aren't worried about suspending other instances of the process by name, you can pass signals to one of these:

killall -STOP program_name
pkill -STOP program_name
pkill -f -STOP program_name_or_args

A CONT signal to a program stopped with Ctrl+z (and not bg'd) will resume its progress (in the foreground), the same as if you were to fg it.

Re: Standard Error

The edit to this question asks about standard error:

The process outputs to stderr, so how shall I issue the command fg <jobid> while the process is outputting to the terminal?

Unless the job in question has components that are backgrounded (or the whole job is backgrounded, perhaps via kill -CONT), you shouldn't actually see output while it is suspended.

If it is still outputting data (be it to standard output or standard error), it will certainly make your terminal visually cluttered, but all of that output will be ignored as it is not part of your input. This might make it harder to know you haven't entered any typos, but (blindly) typing fgEnter should suffice (unless you have multiple jobs and the one in question isn't the most recent, in which case you will indeed need the job descriptor).

If you do need to find the job descriptor, use another terminal to send it the STOP signal via the non-interactive methods above. This should free up your display (perhaps hit Enter a few times or run clear or Ctrl+L) so you can then run jobs to find the job descriptor and then run fg %N where N is that number.

Type fg to bring it to the foreground.