Hello , our CE system is in use

[robot123]

Hello ,

our CE system is in use for automatic scheduling on demand by passing more that 100 schedule tasks every night. The scheduling time is set to "now", so that the jobserver actually should schedule step by step all tasks.
But sometimes the APS forgets one report schduling task and schedules one ore two of all 7 hours to late !)
For example the startTime of one report is
2003.07.30-03:14:21 the jobserver scedules this one at 12 o'clock, where from 4 - 12 o'clock the aps does nothing..
Are there some hidden parameters to optimize mass - scheduling (mass scheduling in a short time range) ??

.. our jobserver scheduls only 2 reports at the same time. Thats full enough, because one report needs not more than 3 minutes and all mass schedule tasks are done after one hour, until one that needs 7 hours to wake up


rgs Andre

 

[Turkbear]

Hi,
Seems a little confusing: How can something be Automatic and On-Demand ?
Why not explicity set the start times for the reports?

 

[robot123]

Im sorry..
the "on demand" featur is not important for this discription. Its so:

a own programm scheduls every night at 3 o'clock a lot of reports using the CE java sdk as following:

ISchedulingInfo schedInfo = myReportInfo.getSchedulingInfo();
schedInfo.setRightNow( true );
schedInfo.setType( CeScheduleType.ONCE );


at 03:04 there are more than 40 reports to schedule. The APS needs about 30 minutes do complete the tasks, but like I said, sometimes some reports have the state "outstanding/queue"
Yes .. the report is is the queue, its starttime is hours in the past and the jobserver does nothing.
When I schedule one more instance in the launchpad of this report, the new instance will be procceded and the old one, too. In this moment the old instance wakes up.
But its not so that the old instance wakes up only in this way. After a long time CE notice by itself that there is actually something to do ..
Then the report will be scheduled at, 4 example, 12o'clock, but thats 9 hours to late..

hope now everything is clear..

rgs Andre

 

[Turkbear]

Hi,
Sorry, I do not use the Java SDK for scheduling, I do it through the CMC ( 8.5 Professional ) , so I don't know enough about that method to see if anything is off..

 

[icytrue]

I have experienced a similar thing to robot123, but when scheduled through eportfolio rather than through code.
Our client has a set of 8 reports that should run at 7:00pm, 7:05pm, 7:10pm etc everyday, these are configured to run with a recurring schedule set up through eportfolio.

About 75% of the time it is fine, but the rest of the time we notice that one or more of the reports runs hours after it should. Sometimes it runs hours late and then fails, but it doesn't always fail (if it does fail it gives a cannot connect to db type error - the whole fail thing might be a red herring - I'm interested in why it does not attempt to run at the appointed time).

I cannot see anything obvious as to why this happens. Has anyone else experienced similar OR even better solved the problem?

thanks


Icytrue

icytrue@hotmail.com

 

[synapsevampire]

It may not attempt to run at the appointed time because CE is busy with other jobs, and the max number of simultaneous jobs has been reached.

Check the reports server properties in the CMC.

Conmsider changing the maximum number of instances (and perhaps some of the maximum threads allowed on the other servers), and check the NT Event logs and the CE logs to determine the cause for failures.

Once you obtain those, if it isn't self evident, post the errors being produced.

-k

 

[icytrue]

we were scheduling for one report to run at a time (or at worst an overlap of say 3 reports running at once). the number of jobs for the job server was set to 10. The cache server and page server (which shouldn't be involved in the schedule process) were set to the default of 75 threads.

I'll get the logs when I next on site!

Icytrue

icytrue@hotmail.com

 

[robot123]

hello,

Ive update the server.report option "parallel jobs" from 2 to 4 jobs.. Now the problem seems to be solved, but we will take a look on it.

Thanks to all involved !

rgs Andre, Berlin