Managing jobs with a queue

Overview

Teaching: 15 min
Exercises: 0 min

Questions

• How can I more easily control many background jobs?

Objectives

• Create a queue of waiting jobs

• Manage a queue of jobs

Overview

Task Spooler (tsp) is not installed on every Linux server. However, we mention it here because if it happens to be available, it is a straightforward way to run many jobs.

Task Spooler is a type of job queuing software. There are many more complicated job queueing software packages. Task Spooler is a particularly simple tool. It only runs on one server for one user.

The tsp command will run a job in the background; this lets you run other commans while tsp is busy running your program. The tsp command will even keep running your jobs after you are logout or disconnect from the server.

In its normal setup, only one job will run with tsp at a time. However, you can ask tsp to run more jobs. Once the first job finishes, tsp will automatically start the next job. This list of waiting jobs is called queuing, and is analogous to the lineup at a grocery store.

Let’s start taking a more concrete look at how to use the tsp command.

Listing tsp jobs

You can view your list of tsp jobs:

$ tsp

ID   State      Output               E-Level  Times(r/u/s)   Command [run=1/1]
1    running    /tmp/ts-out.IiCgZe                           ./run_simulation -t 10
2    queued     (file)                                       ./run_simulation -t 10
0    finished   /tmp/ts-out.QUFZ40   0        2.00/0.00/0.00 ./run_simulation -t 2

The number from the ID column will be used to refer to particular jobs. The State column tells you the progress tsp is making in your job queue.

Running jobs with tsp

Just put tsp in front of the command you want to run. Note that tsp can only directly run a single command, but itself does not handle the special functionality of the Unix shell. So for example if use a command containing pipes or output redirects, then it is easiest to put those shell commands into a file. This file is called a shell script, but is out of the scope of our current discussion.

Let’s again run the run_simulation example program, but this time run it using tsp:

$ tsp ./run_simulation -t 10
$ tsp ./run_simulation -t 3
$ tsp ./run_simulation -t 20
$ tsp

ID   State      Output               E-Level  Times(r/u/s)   Command [run=1/1]
0    running    /tmp/ts-out.Smoz2i                           ./run_simulation -t 10
1    queued     (file)                                       ./run_simulation -t 3
2    queued     (file)                                       ./run_simulation -t 20

We have queued up 3 jobs, each running a run_simulation. We’ll move on now to ways of managing jobs and getting output.

Inspecting results of jobs

Let’s assume you have some finished tsp jobs that look like this:

$ tsp

0    finished   /tmp/ts-out.Smoz2i   0        10.04/0.00/0.00 ./run_simulation -t 10
1    finished   /tmp/ts-out.1UDMiN   0        3.02/0.00/0.00  ./run_simulation -t 3
2    finished   /tmp/ts-out.twnIHU   0        20.11/0.00/0.00 ./run_simulation -t 20

Note that tsp gives each job an ID. You can see the output of a job:

$ tsp -c 1

Using 0 MB of RAM
Using 0 CPU cores

And to see more detailed information about the job:

$ tsp -i 1

Exit status: died with exit code 0
Command: ./run_simulation -t 3
Slots required: 1
Enqueue time: Fri Jul 14 10:45:28 2017
Start time: Fri Jul 14 10:45:38 2017
End time: Fri Jul 14 10:45:41 2017
Time run: 3.021296s

Misc Commands

Remove queued job with ID 2:

$ tsp -r 2

Clear list of finished jobs:

$ tsp -C

Kill a running job with ID 2:

$ tsp -k 2

Using Taskspooler

Let’s assume you have 5 jobs scheduled using tsp. How do you:

1. get more information about the job with ID 2?
2. see the output of job with ID 1?
3. list all scheduled jobs?
4. remove queued job with ID 4?

Solution

1. tsp -i 2
2. tsp -c 1
3. tsp
4. tsp -r 4

Key Points

• The tsp program, if available, will schedule a queue of jobs.