CoCalc Blog

Project Initialization Scripts

Harald Schilly • • cocalc

Under the hood, CoCalc provides you with a very flexible and versatile Linux environment. Due to the upgrade to Kubernetes, our (unofficial) support for using crontab files for periodic tasks was removed. Instead, there is a new and more flexible mechanism to use regular Bash, Python, or SageMath files to accomplish even more.

Project Initialization

When a CoCalc project starts up, an instance of Supervisor is started and responsible for running:

Initialization file management

Supervisor is configured to run this initialization file like background processes. This means it checks the exit value of the process and restarts it if this code is not 0 or 2. Therefore, a script that runs without an error is not restarted.

On the other hand, if the process is terminated, interrupted or crashes – therefore the exit code is e.g. 1 – the process is restarted. This adds some robustness to long running background jobs.

Example 1: record start time

A very simple example is to record the project’s start time. Go to your home directory and create a file with that content:

date > project-start

This is a very simple bash script, which pipes the output of the date command into the file project-start.

In order to see its effect, the file needs to be saved and the project restarted. Give it a few seconds to come back online and run the script. After that (maybe click the refresh button in the file listing) you should see this file and its content might look like:

Mon Sep 12 11:14:20 UTC 2017

Other languages besides Bash?

You can run any language via bash’s exec! For example, containing

exec python3

will run a Python 3 initialization file named

Example 2: a periodic task in Python

Here we write a small Python script, which runs an infinite loop (make sure to use time.sleep!) and evaluates a function every 5 minutes. This examples uses the library schedule. Feel free to choose any other solution.

  1. contains exec python3
  2. The content of is:
import schedule
import time
from random import random
from datetime import datetime

i = 0

def task():
    with open('task_output.log', 'a') as fout:
        ts = str(datetime.utcnow())
        fout.write("Task {}: {} value = {}\n".format(i, ts, random()))
        i += 1


while True:

Indeed, after restarting the project the output of ps auf shows this “daemon” task as a child of supervisor:

  5 ?        Ss     0:01 /usr/bin/python /usr/bin/supervisord -c /cocalc/supervisor/supervisord.conf
 15 ?        Sl     0:01  \_ node /cocalc/src/smc-project/local_hub.js ....
 18 ?        S      0:00  \_ python3 /home/user/

and the output file task_output.log contains:

Task 0: 2017-09-12 12:02:22.636091 value = 0.12071154405652385
Task 1: 2017-09-12 12:07:22.761420 value = 0.6513792691945387
Task 2: 2017-09-12 12:12:22.891285 value = 0.5965113338132986

Example 3: Periodic task in SageMath

run.sage is similar to the Python script above.

  1. exec sage run.sage
  2. This results in Sage running a small task every two minutes and appends outputs to sage_output.log:
import time
from random import random
from datetime import datetime

i = 0
def task():
    global i
    with open('sage_output.log', 'a') as fout:
        ts = str(datetime.utcnow())
        fout.write("Sage Task {}: {} value = {}\n".format(i, ts, random()))
        i += 1

while True:
    time.sleep(2 * 60)


To figure out why a script doesn’t work as it should, there are two ways to debug it:

  1. Run it directly in a terminal (create a *.term file) and run bash or python3
  2. Check the logfile of Supervisor by running this in a terminal: cat /tmp/.cocalc/supervisord.log. Among its logging there are likely entries hinting for exit states (e.g. INFO exited: project_init_sh (exit status 0; expected)) or they show stdout/stderr output of the failed commands.
  3. A common pitfall is to assume ~/.bashrc is run. Since this is a non-interactive session, you need to explicitly source any additional environment information.