Upgrade Planning 7.3 SP3 to MR1

[aptech135]

Has anyone encountered any issues installing 7.3 MR1 for Planning? Biggest concern is admin links and system links still working properly after upgrade.

[andrewbho]

Performance of Admin. Links have been improved.  Any existing links that were created in previous versions will need to be updated.  Any new ones will not have to be rewritten.  Other than that, you can now create application folders and change the administrator of the CAC.  If you haven't upgraded yet, I would upgrade to 8.1.1.

[Smajil]

What is the difference between upgrading to MR1 and 8.1.1?
We have upgraded our DEV environment to MR1 and it went by ok ( Did have to run the configuration manager to refresh the settings)

[ykud]

No big difference, only a few small changes in 7.3 mr1 to 8.1 + contributor data server (for integration with BI) packed inside.
We transfer 7.3 sp3 to 8.1 directly.

[andrewbho]

The BIGGEST difference is the resolution of a memory leak bug that affects the COM services, bug 500972.

If you have a choice, I would upgrade to Cognos Planning 8.1.  If you want to use CDS, you have to install Cognos 8 BI MR2 in order to use it.

[ykud]

The BIGGEST difference is the resolution of a memory leak bug that affects the COM services, bug 500972.

Haven't found that in Cognos 8.1 fix list. Is there any place I can see this bug description.
I've discovered this leaking problems a couple of weeks ago and already found some solution, but I had no idea that bug is registered and solved.

[andrewbho]

memory leaks are very hard to pinpoint as they affect various items.  This specific bug affected the dllhost.  The client was running out of memory.  It is more common in Oracle than in SQL Server.

Something that we just found out that you might be interested in....

Right now, there's an issue in 8.1 with Admin Links.  If you do an internal admin link (source and target are the same) and the elist is matched description, and you target production, the links will fail.  A workaround is to target Development, and then GTP.  There is a hotsite release for this, but you do need to have 8.1 installed.  The bug is also present in MR1.

[ykud]

memory leaks are very hard to pinpoint as they affect various items.  This specific bug affected the dllhost.  The client was running out of memory.  It is more common in Oracle than in SQL Server.

We've got that error in Oracle and in SQL Server. I've searched some workaround solutions and seem to find one. Summed it up in this post. http://ykud.wordpress.com/2006/11/14/memory-leaking-in-contributor/

Something that we just found out that you might be interested in....

Right now, there's an issue in 8.1 with Admin Links.  If you do an internal admin link (source and target are the same) and the elist is matched description, and you target production, the links will fail.  A workaround is to target Development, and then GTP.  There is a hotsite release for this, but you do need to have 8.1 installed.  The bug is also present in MR1.

Thank you for sharing that out. But we've already discovered that. It was in mr1 for sure and got our nerves out.
It's bad, though, that Cognos doesn't allow "open" bug tracking (as metalink, for example), for we could have pointed that bug out before you encountered it.

[andrewbho]

Yep, prior to the fix, we would build and incorporate a VB script to (restart the com+) with our Planning automation.  Before any job would run, it would restart to make sure that we wouldn't have any problems.

Let me know if you want the code.

Void- you should come work for me :)

[ykud]

Yep, prior to the fix, we would build and incorporate a VB script to (restart the com+) with our Planning automation.  Before any job would run, it would restart to make sure that we wouldn't have any problems.

Let me know if you want the code.

Void- you should come work for me :)

Hmm, the code looks like this, or different anyhow? Well -- as far as I see now, there's no need in restarting on Windows 2003 Server, because you have a wonderful option of setting up the COM+ recycling parameter.
I'm from other side of the globe ;)

[andrewbho]

Yep, my script is very similiar but a little cleaner.  We also found it just as effective to create a batch file using net start and net stop against the Cer4 service.  the net result is the same thing but IT folks aren't as intimidated by the code.

[ykud]

Send the script, I'd like to see a cleaner variant.
Restarting service is another variant, but it's considerably slower and, while testing, I've found out that it can take up to 5s on current machine. That's rather near to 15s threshold of main system timeouts. So I've stopped on COM+ restarting.

[andrewbho]

Yep, I will send.  We all won't have to worry about "this" issue soon  ;)