| Yariv,
>Should the T table exist in both servers?
yes
>If yes - why it is created only on one?
on the Commit Changes after the assignment, these tables are created.
I guess soemthing went wrong in the creation process on the management
server. In that case there must be an error reported in the ER-*-* file
somewhere under %OPROOT%\cell.
>If no - how does it work?
In any case - how to procede?
Did you try to restart LNX on the management server? Maybe that helps because
also when a cell is down during the commit, the T.. tables are created at the
first startcell.
If that does not work, you probably have to unassign and reassign the
attributes and Commit the changes.
Jos
|
| I believe that we have discovered why the problem occurred and with the help of
Jos Maandag we have a solution:
We went on-site to take a closer look at things and during our investigation we
discovered an ER-18-97 file from two weeks ago when the SWC was originally
installed. Thie ER-18-97 showed "permission debied" errors on each attempt to
create a T... table. Inspection of the SQL database showed SA as the owner of
the celldb database and NOT omdba. We reset the ownership to be omdba.
Now it all became clear; since the T... tables were NOT created, then an
invalid error would be generated every time a user attempted to create a new
object of this object class. Reinstalling the SWC doesn't solve the problem
immediately since LINKworks thinks that these tables already exist and makes no
attempt to recerate them. This is due to the fact that two system tables,
ATTRTAB & ATTRFELD, contain rows referrencing these T... tables.
The way to solve this problem is to delete the rows which referrence these
tables from both the ATTRTAB and ATTRFELD tables and then "commit changes".
We haven't yet implemented this at the customer site, but we did successfully
test this method on a test SWC which we installed and then dropped the T table.
Thanks to Jos Maandag for his undevoted assistance.
Cb.
|