You are here

Luminis Portal slow/hangs when Banner Database server is down

Submitted by sidman on Thu, 05/10/2012 - 10:16

Hi All,
Our Luminis Portal is extrememly slow and appears to hangs when the Banner Oracle PROD database Host server (not just the instance) is shutdown down.
If we shutdown the Banner Oracle PROD database instance, the Oracle Listener and/or Self-Service Banner on fmw11g all other portal funtionality is unaffected.
Our release levels are:
LP 4.2.1.78
Oracle DB 11.2.0.1.0
fmw 11.1.1.1.4
Banner 8.4
I have an open SR with SungardHE however I've been told "There really isn't much we can do when the backend is yanked from the external systems which will cause a thread pile up....
".
Has anyone else experienced this problem? Have you found a workaround?
Thanks in advance..Sid. Sr Unix Technologist @ U.M.D.N.J.

Luminis Version:

Channels:

Modification:

Comments

I wonder what your relationship is between your Luminis and your
Banner DB.

If you are hosting Luminis DB as a separate schema on the Banner DB,
then yes Luminis does not recover very well (or at all?) from losing its
DB.

If you are calling Banner DB during personDirectory.xml then this is
very different. We did look at making personDirectory respond more
gracefully/timely when one of our external databases was out of action.
This potentially only causes slow down during heavy periods of login.
(Although I happened to check how many POSTs were made to
/cp/home/login during 10 minute intervals across all of our 4 portal tier
webservers, and we had 11000+ POSTs ~17 logins/s )

There are other ways in which the Banner DB may be integrated -
for example you might be using it to perform the authentication for
user logins.

Derek
University of Leeds, UK

Our experience has been that when the OC4J for SSB for SSO has problems, the entire luminis system has problems. Now any time we begin getting DURB timeout errors when launching Luminis in the browser and/or SSB channels not rendering, the first thing we do is have our DBA / SSB server admins restart the OC4J. Typically this resolves the problem within a couple minutes.

Seems to be a design flaw in the architecture. Sungard ... er ... whoever they are now :) ... has acknowledged it and says the Luminis V architecture does not share that achilles' heel.

4.2.0.67

We notice this too, but only when the Banner IP is unreachable. If the oracle listener / the Banner db is down, but the server can still be ping'd by Luminis, it seems to be fine.

We have our Banner database on a seperate server from the Luminis Portal database. We have the same experience as Derek in that the portal hangs only when the Banner IP is unreachable.
Unfortunately, per SungardHE this is the expected behavior and the only workaround is to have the Banner Database in a Oracle cluster such as RAC or a Dataguard scenerio. This is not happening right now.
I guess we'll have to live with it.
Thanks for the feedback...Sid.

For extended banner outages, what we do is to put an entry into the OS hosts files to trick luminis into thinking that banner is some other IP. You could probably even use 127.0.0.1

127.0.0.1 localhost bannerName bannerName.domain.edu

However, we also update dns to reflect the hosts file changes. Our Luminis env. has its own dns (done for various reasons). So I'm not 100% sure that hosts files are enough.