What's changed: Initial version (topic 1.01, subtopics 1.01.1–1.01.5)
1.4Creating, Monitoring, and Terminating Processes
Learn basic process management: running jobs in the foreground/background (&, bg, fg, jobs), keeping work running after logout with nohup/screen/tmux, monitoring with top/ps/pstree/uptime/pgrep, and sending signals with kill/pkill/killall.
Run a long task while doing other work, stop a runaway process, find what is loading the box—server operations are daily process management. Master shell job control, signals, and the monitoring commands.
1.4.1Job control and surviving logout
- Append & to start in the background. Ctrl+Z suspends; resume with bg or bring back with fg. List with jobs.
- nohup keeps a command running after logout (terminal hangup = SIGHUP). The idiom is
nohup long-task &. - screen and tmux are terminal multiplexers: detach/reattach sessions at will, so work survives a dropped SSH connection.
1.4.2Monitoring and terminating with signals
- Monitor with top (live CPU/memory leaders), ps (snapshot;
ps auxis the staple), pstree (parent-child tree), and uptime (uptime and load average). - pgrep finds PIDs by name and attributes (
pgrep -u user01 nginx). - Terminate with kill (by PID), pkill (by name/attributes), killall (exact name). Default is SIGTERM (15, polite request); the last resort is SIGKILL (9, force). SIGHUP (1) means hangup / reload config.
Staples: survive logout = nohup (or screen/tmux), suspended → background = Ctrl+Z then bg, SIGTERM first, SIGKILL only if needed, find by name = pgrep/pkill, send by PID = kill. Jumping straight to kill -9 is rarely the right answer (no cleanup runs).
Walk the typical scenarios. Start long builds with nohup make all > build.log & and they survive dropped SSH. For more flexibility open a tmux session, detach with Ctrl+b d, and tmux attach later. If you suspended an editor with Ctrl+Z, check jobs and bring it back with fg %1 or push it back with bg %1. For load hunting: uptime for the load average trend, top for CPU/memory leaders, pstree to see which parent spawned it. To stop something, confirm the PID with pgrep -f app, then escalate: kill <PID> (SIGTERM) → kill -9 <PID> (SIGKILL) only if it will not die. Note killall signals every process whose name matches exactly—it can hit unintended same-name processes.
| Goal | Use | Note |
|---|---|---|
| Keep running after logout | nohup / screen / tmux | screen/tmux can reattach |
| Suspend → resume in background | Ctrl+Z → bg | fg brings it forward |
| Load overview | uptime / top | Load average → top processes |
| Graceful stop → force | kill (SIGTERM) → kill -9 (SIGKILL) | Avoid jumping to -9 |
Trap: "kill terminates a process by name" is wrong—kill takes a PID (by name is pkill/killall). And "SIGKILL (9) lets the process clean up" is wrong—SIGKILL is uncatchable and immediate; it is SIGTERM that allows cleanup.
1.4.3Section summary
- Job control = & / Ctrl+Z / bg / fg / jobs; survive logout with nohup, screen, tmux
- Monitor = top/ps/pstree/uptime/pgrep; terminate SIGTERM → SIGKILL, by name with pkill/killall
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. You want a batch job to keep running even if the SSH connection drops. Which command?
Q2. You pressed Ctrl+Z and suspended a job. Which command resumes it in the background?
Q3. A process (PID 4321) is unresponsive. What should you try first?