I want to run a job every 10 minutes,
but at times the job could run for 20 or 30 minutes.
How can I ensure that the new job does not start
up again?

Convert it to a script type job. Use a global variable that you set/check in the beginning of the script and rest at the end. If the variable is set it means another job instance is running so that the current instance should simply exit without doing anything.
Another solution is to make job synchronous and assign it to a separate dedicated job queue. This will ensure that only one job instance can run at a time.

: I want to run a job every 10 minutes,
: but at times the job could run for 20 or 30 minutes.
: How can I ensure that the new job does not start
: up again?

I tried both of these options and what happens is the job stack up so when the
first job finishes then all the other jobs that did not run execute one after the other..

: Convert it to a script type job. Use a global variable that you set/check in
: the beginning of the script and rest at the end. If the variable is set it
: means another job instance is running so that the current instance should
: simply exit without doing anything.
: Another solution is to make job synchronous and assign it to a separate
: dedicated job queue. This will ensure that only one job instance can run
: at a time.

That's right. That is exactly what should happen if you use the second method,... and you get only one instance of the job running at a time. If there is nothing to do the job should simply exit without doig really anything. Add "to do" check to the job logic and you should be fine.

If you decide to stick wth the first method, make sure the job is set to run async. Also don't confuse job start-end events in the log with job really doing the work. The log simply indicates that the job wanted to start. However if implemented correctly the job must check the global flag and exit immediatelly after start if it finds another instance of the job already running.

: I tried both of these options and what happens is the job stack up so when
: the
: first job finishes then all the other jobs that did not run execute one after
: the other..

Let me clariy, I do not want the job to run until the next scheduled time.

: That's right. That is exactly what should happen if you use the second
: method,... and you get only one instance of the job running at a time. If
: there is nothing to do the job should simply exit without doig really
: anything. Add "to do" check to the job logic and you should be
: fine.

: If you decide to stick wth the first method, make sure the job is set to run
: async. Also don't confuse job start-end events in the log with job really
: doing the work. The log simply indicates that the job wanted to start.
: However if implemented correctly the job must check the global flag and
: exit immediatelly after start if it finds another instance of the job
: already running.

Use method 1. If you don't want to see job related start/end events in the log, uncheck the logging option and handle logging in the script using LogAddMessageEx. Call this only if the initial check finds no other job instances.

: Let me clariy, I do not want the job to run until the next scheduled time.