Y2K test machine deadlocks stop our project

[Michael Lee]

Hi Andreas, we too have been experiencing several problems in upgrading to 819C.
Lately they're mostly performance problems. With Cognos's help we have found
that there is quite a major memory leak in the 819C4 (819C2, 819C3 included)
version of Powerhouse (with Allbase). Cognos has identified a lot of the
problems and fixed them in the 819C5 (not yet out) version. However they still
have a ways to go. We tested a version of 819C5 and found that it still ate up a
lot of memory-just slower than 819C4.

It looks like we might be using the 729C9 version of Powerhouse instead. This is
a real pain because the CENTURY WINDOWing option was extremely useful. In
retrospect it would have been nice to have changed all our dates to FORMAT
YYYYMMDD, but since we didn't we have had to hard code CENTURY WINDOWing in the
EDIT procedure for all our date fields (luckily we had USE files for most of our
dates already in place). However much to my disdain I found that the EDIT
procedures don't get executed by the REQUEST verb (in the PATH procedure for
example). As soon as it happened I realized it was something I should have
known. This means we had to change those dates that are prompted for during a
find (about 10-15 screens).

Cognos has gone to great lengths to help us (kudos to Doug Clement, Bob Deskin,
Mark Munday, Steward Rich, Peter Bateman et al; and probably several more people
I forgot), but are having a hard time nailing down all the issues. It's quite a
complex change that needs to be done. We are still testing the 729C9 version and
have found a few idiosyncracies some of which you wouldn't believe if I told
you. Most of these go away if I recompile the program (?!?) in question however.

Feel free to contact me directly via e-mail if you want to share information
Andreas. If we can't help each other at least we could commiserate together.

Michael Lee
MCL Systems Inc.


Noack, Andreas M. wrote:

> Dear PowerHousers,
>
> as I already proposed sometime earlier this year let me give some status
> update on our test installation. When pushing our application to PH V 8.19
> C4 on HP3000/MPEiX 6.0 with Allbase G3.02 we were experiencing (among other
> ) locking problems in QTPs. With some detective work we found the following
> bloopers, and that's just mentioning the "big" ones:
>
> - when having no "run <modulename>" in the QTPs they tend to lock each other
> -> workaround: add this statement to all QTPs
> -step 1 of QTP accesses table, step 2 just a subfile, step 3 again a table,
> then no correct module created (only steps 2,3), LOCK -> workaround: add a
> dummy link to table in step 2 of all QTPs like this
> -problems with viaindex -> workaround: use explicit link
> -item yearmonth, int siz 4, pic MM/YY could not be accessed anymore ->
> workaround un- and reload all tables containing this
>
> So you see, in contact with Cognos telesupport we were rather proactive and
> successful. But the current situation leaves us baffled again: we had 4
> users on our test machine in 2 environments to simulate "real life"
> processing. The test machine is an HP3000 918 LX with 256 MB main memory, 1
> processor and 10 GB HDD, very full. So those poor users ran firstly into big
> performance problems (that was to be expected) and created about 37
> deadlocks in two days with QTPs already overworked and tested (that one was
> a hard hit!). Now we are suspecting something like the size and performance
> of this machine to be the reason for our trouble. As this is just a guess
> and hiring a big machine just for testing would be pushing extremly on our
> budget can anybody state similar experience from his own machinery (maybe
> also on other platforms). As this is rather urgent by now the question went
> to telesupport, too, while I am posting to the list. Any hints whether we
> may be guessing right, also from the Cognos pros over here, will be greatly
> appreciated (know you can do it, folks). THX in advance for help of any
> kind.
>
> Kind regards
> Andreas Noack
> BetzDearborn
> A division of Hercules
> Development Group Europe
> Tel. : +32.16.395.290
> Fax : +32.16.395.368
>
> The opinions expressed above are my very own and not those of my
> employer. Due to this fact they may change anytime without further
> notice.
>
> = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
> Subscribe: "subscribe powerh-l" in message body to majordomo@lists.swau.edu
> Unsubscribe: "unsubscribe powerh-l" in message to majordomo@lists.swau.edu
> powerh-l@lists.swau.edu is gatewayed one-way to bit.listserv.powerh-l
> This list is closed, thus to post to the list, you must be a subscriber.



= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Subscribe: "subscribe powerh-l" in message body to majordomo@lists.swau.edu
Unsubscribe: "unsubscribe powerh-l" in message to majordomo@lists.swau.edu
powerh-l@lists.swau.edu is gatewayed one-way to bit.listserv.powerh-l
This list is closed, thus to post to the list, you must be a subscriber.