Scheduler - Problems night time

Hi,
we've a problem with 24x7 Scheduler.
It happens sometimes that, also if the scheduler is running and ther aren't message errors,
no job start and the timer giving remaining time before the next job seems to be wrong
(For e.g. this morning, timer was signaling us that the next job would be started in 2 hours, when actually the next job had to run in 15 miuntes)
Usually the scheduler stops working night time, so we notice the problem the morning after.
Only after a stop and start of the scheduler, it returns working.
Analyzing the various job logs and the windows event viewer no such information are given us about errors.
I've already scheduled, in the Options Menu, an autorestart of the program every day ad 04.00 A.M.

The scheduler runs over a Windows 2003 Server clustered machine.

Have you got any suggestion about this problem ?

Thanks in advance
Stefano Pantaleo

Hi,

The countdown clock shows when the next job is ready to be launched. Please note that the job goes to the queue first and waits in the queue until all previously queued jobs finish executing, I mean synchronous jobs in the same job queue. If you see some job running it doesn't mean that the running job is that the one counted by the countdown clock. It could be a job who was waiting in the queue and whose turn came to run. It could be al so an event-based job, such as a job with file watch, process watch or email watch schedule type.

If you rest mouse pointer over the countdown clock for a couple of seconds, it will show which job it considers ready to launch next.

As general troubleshooting steps, do the following:
1. Ensure "Show job status display" option is turned off in the system settings.
2. Jobs for different applications are assigned to different job queues (use job queue property to assign job to a specific queue)
3. Total number of job queues should be in 1 to 10 range.
4. Jobs should be set to run detached and synchronous (use corresponding job properties.)

SysOp wrote:

Hi,

The countdown clock shows when the next job is ready to be launched. Please note that the job goes to the queue first and waits in the queue until all previously queued jobs finish executing, I mean synchronous jobs in the same job queue. If you see some job running it doesn't mean that the running job is that the one counted by the countdown clock. It could be a job who was waiting in the queue and whose turn came to run. It could be al so an event-based job, such as a job with file watch, process watch or email watch schedule type.

If you rest mouse pointer over the countdown clock for a couple of seconds, it will show which job it considers ready to launch next.

As general troubleshooting steps, do the following:
1. Ensure "Show job status display" option is turned off in the system settings.
2. Jobs for different applications are assigned to different job queues (use job queue property to assign job to a specific queue)
3. Total number of job queues should be in 1 to 10 range.
4. Jobs should be set to run detached and synchronous (use corresponding job properties.)

Hi,
Now we've understand how to see what is the next job ready to run, and if has happened again, we would check if it is that block all the queues.
But there's nothing we can investigate to better know what's happened, and prevent it would happen again?
Regarding the 4 points to check, we respect "almost" all of them, a dubt about #3.. what do you mean with "Total number of job queues should be in 1 to 10 range." ? Because we have almost 50 queues, one per program need to be scheduled.
Notice that usually the scheduler works well; the problem mentioned above raises about every 2 weeks,s o we can't be sure that the problem is caused by that options.

Stefano Pantaleo

Just as a comment, using 50 queues seems to be overkill. You are stressing out the entire system, not just the scheduler. Queues are designed to organize jobs into controlled job streams and parallelize the processing. You are likely using a separate queue for each job. It is unlikely that all 50 jobs would ever need to run at the same time. Right?

	SysOp wrote:
	Just as a comment, using 50 queues seems to be overkill. You are stressing out the entire system, not just the scheduler. Queues are designed to organize jobs into controlled job streams and parallelize the processing. You are likely using a separate queue for each job. It is unlikely that all 50 jobs would ever need to run at the same time. Right?

Hi,
I think there is a misunderstanding in our posts and replies
So I'll try to better explain our situation.
We've configured a scheduler server to manage, in one place all the programs that have to run on remote computers at precise time.
By now we've configured about 50 jobs (many others will be added) that run on different computers almost never at the same time.If we consider the 24 hours, during a day it will be an average of about 2 jobs per hour.
Maybe could happen that some jobs run simultaneously but this is not the rule and in that case they are never more than 3 or 4.If we had known the limitation of 10 job a day (as we could suppose from your last replay) we wouldn't had buy a professional scheduler as we thinked 24x7 was, but instead we would used the windows scheduled tasks service.
In the documentation we didn't find this limitation mentioned.

About the queues.
We have "built" a queue for each "logical job" needs to run and setted to be notified about failed jobs via mail or windows message. What happen instead is that in the morning we notice that none of the scheduled jobs have run during the night and the scheduler didn't signal it to us standing stopped without any message.
We are asking you if there is any log, or something else in the scheduler to look at,to understand the why his behaviour.
Could you help us to find out a solution for this problem ?

Thanks in advance

Stefano Pantaleo

There is no 10 job limitation. You can run virtually as many jobs as many jobs your computer can handle.

I believe you misunderstood how queues work.

Consider this hypothetical example: you edit only 10 small Microsoft Word documents a day for 1 minute each. Yet you keep 50 instances of Microsoft Word program running on your Desktop all the time, for the only reason "what if you need to edit 50 documents at the same time for 1 minute each" Documents can be stored on a local drive or on a network share. While you can still use your Desktop, having 50 Words open all the time is a bit inconvenient. Isn't it? Not to mention these 50 instances require lots of memory and cause the system to do memory swapping to disk, use additional CPU ticks, etc… making everything to run slow.

I suppose in your situation, 2 or 3 queues would do just as well as 50 or 100. Please note that job queuing has absolutely nothing to do with job notifications or job placements (local/remote). Neither number of queues directly controls how many jobs can be scheduled or run. Again, consider each job queue as a separate independent and concurrent scheduling system.

I suggest that you reduce the current number of queues to a meaningful number. Spread all 50 jobs across the available queues. Jobs that take hours to run, should be set to run asynchronous so that they don't block their queue. Jobs that run for minutes or seconds you should keep synchronous.

Clear the job log. You can analyze it in a few days and check if the proposed configuration is stable and if you have any job run-time dependencies. For that there is also job timing and conflict report, but for the report to work you would need to enable job performance statistics option.