Run CRON job every 5 minutes on OpenShift (Red Hat

问题:

I'm trying to run this script every 5 minutes. It seems the only way to run CRON jobs on OpenShift is to use their CRON plugin. And the CRON plugin only allows for minutely, hourly, and daily scripts (by placing the script in the corresponding folder).

I am trying to run this script every 5 minutes:

#!/bin/bash
php /var/lib/openshift/53434c795973ca1ddc000668/app-root/runtime/repo/scheduled.php > /dev/null 2>&1

But right now it runs every minute (because it's placed in the minutely folder).

How can I re-write it so that it runs every 5 minutes?

回答1:

Modify the script so it checks the current time, and bails out if it's not a multiple of 5 minutes.

Something like this:

#!/bin/bash

minute=$(date +%M)
if [[ $minute =~ [05]$ ]]; then
    php ...
fi

The right operand of the =~ operator is a regular expression; the above matches if the current minute ends in 0 or 5. Several other approaches are possible:

if [[ $minute =~ .[05] ]]; then

(check for any character followed by a 0 or 5; $minute is always exactly 2 characters).

(User theshadowmonkey suggests in a comment:

if [ $(($minute % 5)) -eq 0 ]; then

which checks arithmetically whether $minute is a multiple of 5, but there's a problem with that. In the expression in a $(( ... )) expression, constants with leading zeros are treated as octal; if it's currently 8 or 9 minutes after the hour, the constant 08 or 09 is an error. You could work around this with sed, but it's probably not worthwhile given that there are other solutions.)

回答2:

I will extend Keith Thompson answer:

His solution works perfectly for every 5 minutes but won't work for, let's say, every 13 minutes; if we use $minutes % 13 we get this schedule:

5:13
5:26
5:30
5:52
6:00 because 0%13 is 0
6:13
...

I'm sure you notice the issue. We can achieve any frequency if we count the minutes(, hours, days, or weeks) since Epoch:

#!/bin/bash

minutesSinceEpoch=$(($(date +'%s / 60')))

if [[ $(($minutesSinceEpoch % 13)) -eq 0 ]]; then
    php [...]
fi

date(1) returns current date, we format it as seconds since Epoch (%s) and then we do basic maths:

# .---------------------- bash command substitution
# |.--------------------- bash arithmetic expansion
# || .------------------- bash command substitution
# || |  .---------------- date command
# || |  |   .------------ FORMAT argument
# || |  |   |      .----- formula to calculate minutes/hours/days/etc is included into the format string passed to date command
# || |  |   |      |
# ** *  *   *      * 
  $(($(date +'%s / 60')))
# * *  ---------------
# | |        | 
# | |        ·----------- date should result in something like "1438390397 / 60"
# | ·-------------------- it gets evaluated as an expression. (the maths)
# ·---------------------- and we can store it

And you may use this approach with hourly, daily, or monthly cron jobs on OpenShift:

#!/bin/bash
# We can get the

minutes=$(($(date +'%s / 60')))
hours=$(($(date +'%s / 60 / 60')))
days=$(($(date +'%s / 60 / 60 / 24')))
weeks=$(($(date +'%s / 60 / 60 / 24 / 7')))

# or even

moons=$(($(date +'%s / 60 / 60 / 24 / 656')))

# passed since Epoch and define a frequency
# let's say, every 7 hours

if [[ $(($hours % 7)) -ne 0 ]]; then
    exit 0
fi

# and your actual script starts here

Notice that I used the -ne (not equal) operator to exit the script instead of using the -eq (equal) operator to wrap the script into the IF construction; I find it handy.

And remember to use the proper .openshift/cron/{minutely,hourly,daily,weekly,monthly}/ folder for your frequency.

回答3:

You can make use of minutely cron job available. The jobs are run at a specified frequency and you can instrument your job to inspect the date and/or time when your job runs.

Here is an example for running a job every 5 minutes. Place the below code snippet in an executable file at path .openshift/cron/minutely/awesome_job and give proper permissions chmod +x awesome_job then add it to your application repository, commit and push.

The snippet

#!/bin/bash
if [ ! -f $OPENSHIFT_DATA_DIR/last_run ]; then
  touch $OPENSHIFT_DATA_DIR/last_run
fi
if [[ $(find $OPENSHIFT_DATA_DIR/last_run -mmin +4) ]]; then #run every 5 mins
  rm -f $OPENSHIFT_DATA_DIR/last_run
  touch $OPENSHIFT_DATA_DIR/last_run
  # The 'awesome_command(s)' that you want to run every 5 minutes
fi

Explanation:

• This snippet actually creates an empty file named 'last_run' just to note down the last run status and then starts your job.
• The next minute, your 'awesome_job' will be executed and now it tries to find file 'last_run' modified in last 4 minutes. But your file 'last_run' was created just a minute ago and hence the condition fails and exits. This continues.
• At fifth minute, while finding (with -mmin +4), your 'last_run' will be shown and now, the file 'last_run' will be deleted and recreated and your awesome command(s) will be executed.

Source (Official Docs): https://developers.openshift.com/managing-your-applications/background-jobs.html#_execution_timing