I defined the "d" queue for runing only one job at a time and I need to stop this queue for make backup of the oracle database. I stop the queue modifying the line of queuedefs to "d.0j0n1w". I stream jobs that I need to run after the backup finalizes and I need jobs run in order I streamed them, but, when I restart the queue, the jobs run in aleatory order.
Refering to the wrong order: When queue is stopped I stream jobs that execute large cobol programs that I need run at night with nobody working. Every cobol program needs to run in order it has been streamed. If I stream first a job that runs the program called A and after this I stream another job that runs program B, and some user streams a job for program C I need they run in order they has been streamd beacause program B needs that program A has finished for make their job and finally can run the user program C.
Sorry for my bad explain of the problem. Thanks for your interest.
The message I recive after queue jobs is like this:
From dimpsg@itaca.itaca.mgc.es Thu Jun 9 00:01:02 MET 2005 Received: (from dimpsg@localhost) by itaca.itaca.mgc.es (8.11.1/8.11.1) id j58M12B19139 for dimpsg; Thu, 9 Jun 2005 00:01:02 +0200 (METDST) Date: Thu, 9 Jun 2005 00:01:02 +0200 (METDST) From: dimpsg@itaca.itaca.mgc.es Message-Id: <200506082201.j58M12B19139@itaca.itaca.mgc.es>
warning: commands will be executed using /usr/bin/sh job 1118268067.d at Thu Jun 9 00:01:02 2005
$ at -l -q d 1118353786.d Thu Jun 9 23:49:46 2005 1118353788.d Thu Jun 9 23:49:48 2005 1118353789.d Thu Jun 9 23:49:49 2005 1118353790.d Thu Jun 9 23:49:50 2005 1118353791.d Thu Jun 9 23:49:51 2005
How long all these jobs run?? For all those scripts, at the start can you do date >> some log??
We will verify what exactly is happening. I am doubting that the time difference is so small that it might appear that they are being executed in wrong order.
I think at is doing what it is supposed to do. The time difference between jobs is so small that you won't notice anything. At is used to start those jobs at defined times and not to start second one after one starts.
If you want job2 to start after job1 ends, then you need to take care of it. Put a condition in job1 something like follows.
The end of the job1 would look like follows.
jobe1 code xxx xxx status_job1=$? if [[ status_job1 -eq 0 ]] then at -f "job2_script" -t now +1 fi
But I don't know "job2_script" beacause it's can be queued by any uset at any time from interactive programs and I can not include it in "job1".
Now I have an idea, if at end of every job I add the execution of a script like this:
$cat personal_rescheduling jobs_dir="/var/spool/cron/atjobs/" ls -1 $jobs_dir | while read num_job do at -qd now < $jobs_dir$num_job at -r $num_job done
I think all will be ok, what do you think about this ?
That is what I am advising you to do. What you need to do is job1 ends and starts up scond job and second job ends and starts up third.
You not only get the correct order in which they would start, you can also put error code checking to decide to start next job or not, OR to start in different way depending upon error code.
Sorry, now I think this is not completly correct. If I stop the queue, when I restart it, first job to run can be any of queued jobs. After this first job runs, all other jobs run in correct order, but I don't know what job will be the first.
Except if in the script that change queue limit from 0 to 1 and viceversa I append at the end same personal_reschedule script.
At this point, I think only one little problem can have, every time I reschedule a queued job, the at command inserts at the begining of the script all of the environment variables. Is this a serious problem or I can discard it ?
Sorry for don't undrestand what you try to tell me in your last message, sorry.
