Same EXTERNAL DEPENDENCIE resolve for two Ocurrence added

dbordon · New User Joined: 25 Jan 2007 Posts: 20 Location: Argentina

Could you help me figure out the following:
I had three applications which are adding trought batch job (not always using ETT) application xxxxx into the Current plan.

Problem is that two ocurrences of applid xxxxx runs almost togheter.
How this happened?

See OPC audit:
05/02 23.05.12 CP UPDT BY TESTER MCP ADD APPL: XXXXX APPL: XXXXX IA: 120502 2303 PRTY: 5
- OPNO:015 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:009
- OPNO:020 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:015
- OPNO:030 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:020
- OPNO:035 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:030
- OPNO:089 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:035
- OPNO:009 TYPE: EXT. DEP ADD TYPE: PRED OPNO:089 APPL: XXXXX 120502 2232
05/02 23.05.12 CP UPDT BY TESTER MCP ADD APPL: XXXXX APPL: XXXXX IA: 120502 2300 PRTY: 5
- OPNO:015 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:009
- OPNO:020 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:015
- OPNO:030 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:020
- OPNO:035 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:030
- OPNO:089 TYPE: INT. DEP. ADDED TYPE:INT. DEP. ADDED TYPE: PRED OPNO:035
- OPNO:009 TYPE: EXT. DEP ADD TYPE: PRED OPNO:089 APPL: XXXXX 120502 2232

Applid XXXXX runs "toghether" becuase they acknowledge the same EXTERNAL DEPENDENCIE IATIME:
- OPNO:009 TYPE: EXT. DEP ADD TYPE: PRED OPNO:089 APPL: XXXXX 120502 2232

I cannot understand what happened. Maybe i should consult IBM.

David Robinson · Active User Joined: 21 Dec 2011 Posts: 199 Location: UK

So are these two applications both being added in one step, but you want one to be dependent on the other? I assume that's the case as the audit timestamp is 23:05:12 for both.

Which app are you adding first? From the audit it looks as though the 23:03 version is added first. This applies the dependency on the previous 22:32 schedule which is presumably correct. Then when you add the 23:00 version, it will apply the same dependency as that is the only application with an earlier or equal IA time.

If you added the 23:00 version first, followed by the 23:03 version, you may get different results.

dbordon · New User Joined: 25 Jan 2007 Posts: 20 Location: Argentina

Thanks for your interest David.
The thing is like this:
Applications APPL1, APPL2 and APPL3 are adding application XXXXX into the current plan.

Some of those three applications are triggered by an ETT, other no. So application XXXXX can be added to the current plan at any time.

Now, in this particular case application XXXXX is was added twice at the same time, by two of those applications (APPL1-2-3).

TWS assign a different "start" IATIME for each of these new Current Plan ocurrences, but why it assigns the same EXTERNAL IATIME?
When program do not assign different EXTERNAL IATIME for these equals ocurrences both of them were allow to start.
Obviously they made a little disaster by executing almost at the same time by JES2, the operations were mixed and came up with a DISP Dataset invalid (yes! because one previous operation deleted a dataset that the next one use).

At that time is probably that system was heavily used, but TWS mix up the external IATIME, i think that was not possible.

My question is: Should i think that it is something wrong with TWS?

For last, TWS audit is helpful but there is no documentation to support it. I know it is not "rocket science" but... for example the timestamp of the begginng of each record correspond to the order of process by JT VSAM?

David Robinson · Active User Joined: 21 Dec 2011 Posts: 199 Location: UK

I'm still not quite clear on this. Isn't the external IA time that you refer to simply the IA time of the earlier occurence, ie; 22:32? There is no "External IA time" as such. The EXT DEP ADD message is simply showing you which occurence of the earlier schedule the dependency is being resolved to, and there is no reason why that shouldn't be both the same one.

This is what I would expect with the IA times of the new applications that are being added, ie; 23:03 followed by 23:00. How are these IA times generated? Are these specified by you or defaulted by TWS? If the latter I don't see how two applications added at 23:05 can have IA times of 23:03 and 23:00. They should be 23:05 and 23:06.

The examples you show above are not being added by ETT, if they were the audit would say UPDT BY ETT. Presumably TESTER is a batch or TSO user?

Finally, for this scenario, I would suggest special resources might be a better way of controllong the schedules. Assuming you are on v8.3 or later, have the first job in the schedule use a special resource which gets set to unavailable when the job completes. The last job could then issue a SRSTAT to set it back to available.