Publisher Jobs re-initiate (restart) on different Execution Service

[#1]

We continue to experience production load failures due to Publisher re-initiating jobs on different Execution Services (different servers in our case).

· We are running Publisher 8.5 which includes four execution services, each of which reside on separate yet identical servers (128GB memory, four Quad-Core processors, 300GB harddrive + attached SAN of 250GB)

Description of problem:

· QV Publisher runs a daily job as scheduled or initiated by EDX. Immediately following successful completion of the job, Publisher initiates the job again on a separate server (different Execution service). The job that runs the second time typically fails or causes contention with other running jobs.

· This problem has been noted since 2009 and has affected several different jobs.

· All the 're-initiating' jobs are EDX enabled.

· All the jobs have the "Attempts" set to 0 (setting found in Job setup in Publisher > Advanced > Attempts)

· Example: Job "Outfitter_Sales" ran this morning at 5:07am on Server 021 finishing successfully at 5:16:46 am. 13 seconds later the job re-initiated (at 5:16:59am) on Server 023 and caused job failures and a phone call from the computer room.

First run, successful, on server usap021 (the 'primary' execution service); Second, re-initiated, run on server usap023.

Any help from anyone on this?!

Thanks! Gracias!

[#2]

Here's the disappointing news from qlikview support.

Resolution: We have reconfirmed with engineering that everything is working as; designed here. EDX was never intended to be used in a multi-Execution; Service environment as far as 8.xx goes.; That being said, let me try to explain things here. When too many EDX; requests flood Publisher, and they queue up, the XS's will lose the file; lock on the QVW at some point if it cannot assign a QVB quickly enough.; In a normal publisher job in 8.xx, this included a 5 minute timeout; setting on the QVB request, so if a job attempted to run, and all QVBs; were in use, publisher would run a timer for 5 minutes on the job, and; if no QVB had freed up to take the job, the job would be killed.; When running things via EDX, this timer is not involved, so what happens; is the EDX ends up hanging for all intents and purposes, and when you; manually rerun things, this seems to reset things and then we believe; the Number of Retries setting on the Job is likely kicking in and; causing the extra run of the job, so the workarounds are to be sure you; have enough QVBs available to handle all of the EDX tasks coming in and; that you set the Number of Retries on the job to 0, which should; alleviate the extra run you are encountering after running manually.

[#3]

It seems this re-initation or execution of jobs in Publisher (8.5) is getting progressively worse now. We have taken all the precautions recommended by QV support, but to no avail.

My questions to anyone out there:

Has this error been addressed in QV 10?

What is the role, if any, of the "Retry Delay Seconds" setting in the Job setup on Publisher? It defaults to 60 seconds... Does anyone know what would happen if that were changed to, say, 1,000,000 seconds?

Thanks / Gracias!