TM1 Forum

Folks

I have an unusual situation in that I have a chore that runs hourly. The chore runs a single process and the setting is such that it treats the chore as a single transaction. The single process has about 8 sub processes and takes about 2 mins to run.

There is no start batch updates / end batch updates logic within any of our processes.

The server is set up for PI but we don't carry out view creation in separate processes thus a process will lock out others for its duration rather than for the limited duration of the process that builds the view.

Without contention the chore runs fine, but when a process locks it out, it seems that the scheduler will resubmit the chore again and again so it runs many times afterwards locking any other programs out.

Note: the multiple submissions are not as a result of previous submissions unable to run being queued up. Before the problem the previous hour's chore would have run successfully.

Any ideas would be much appreciated. Happy to provide any further information.

Regards

John

First, welcome back. As far as I'm aware it's been a few years since you darkened the doorstep.

Second, could you please advise which version you're on now? Your profile still says 9.0, which seems unlikely.

Hi Alan

Thanks for the update. Yes, it has been a long time - think you had the Mr T icon back then! I've updated my profile as requested.

Absence down to overzealous internet filters. Many times I tried to point out that this was an aid to TM1'ers only to be met with a sea of indifference and beauraucrasy (never can spell that word). Not quite sure what you can do to combat that, but the intellectual property right notice on the forum goes a fair way to ensuring that institutions don't throw a wobbly when you request site access.

Thanks for keeping the site going. I have done a fair bit with PM and CAFE now and I will try and post my experiences for the wider community.

Look forwards to chatting to you all again.

Regards

John

It's PAx now John not CAFE.....

Sounds to me that your chore is getting stuck in rollback loop that is clashing with the other jobs lock.

You can turn the logging up so that you can see what is happening in the locking model, this might give you some clues as to what is really happening. (https://www.ibm.com/support/knowledgece ... 6050c.html)

The simple thing to do is probably just put some stop start flags up so that the chore exits on line 1 if the other jobs are running, this would mean that the chore is not run though which may not work for you.

If you really need the chore to run every hour without fail then maybe
Put some flagging in the other TIs that mean they exit if the chore is running or if it is close to the time window the chore is running in in. I.e. figure a way out to give the core priority on the locking model.
Upgrade (not normally practical I know).
Rewrite to avoid the contention, might not be possible depending on the release.

Cheers,
Steve

as written above the use of flags is a good idea.

I'll suggest an alternative to file flags:

Use your chore to run one parent process that starts the child processes.
in the parent process in the first line of the prolog, use the function