SEO

Platform: Solaris 9, Oracle EE 8.1.7.4, Oracle EE 10.2.0.1

Two databases on the same box, call them eps8pd (8.1.7.4) and epspd
(10.2.0.1)

In eps8pd, I created a db link pointing to epspd:

CREATE PUBLIC DATABASE LINK "LNK_EPSPD" CONNECT TO "NOBELEPS"
IDENTIFIED BY "********"
USING 'EPSPD'

then, connected to eps8pd as NOBELEPS (same user exists on both db's):

Connected to:
Oracle8i Enterprise Edition Release 8.1.7.4.0 - Production
With the Partitioning option
JServer Release 8.1.7.4.0 - Production

SQL> select count(*) from user_tables@lnk_epspd;
select count(*) from user_tables@lnk_epspd
*
ERROR at line 1:
ORA-02068: following severe error from LNK_EPSPD
ORA-00603: ORACLE server session terminated by fatal error
ORA-00604: error occurred at recursive SQL level 52
ORA-00036: maximum number of recursive SQL levels (50) exceeded
ORA-00604: error occurred at recursive SQL level 52
ORA-00036: maximum number of recursive SQL levels (50) exceeded


Searching the ng archives, MetaLink, and Google on ORA-00036 gives a
few hits, all talking about the likelyhood of a recursive trigger. But
as you can see, my example is a simple select that shouldn't generate
much recursive SQL. Linking FROM the 10.2 db TO the 8.1 db causes no
problems. I realize that this may be an un-resolvable problem with 8
-> 10 connections, but am making one last attempt to find a possible
work-around before going to plan 'D'.

Thanks.

Kind of a reach, but are you tracing?
https://metalink.oracle.com/metalink...,1,1,helvetica

Just for jollies, check MAX_TRANSACTION_BRANCHES and
DISTRIBUTED_TRANSACTIONS on the 8 side (though it looks like the 10
side is throwing the error - are you sure it doesn't have more links?).

I think you are going to have to have a 9 in between.

jg
--
@home.com is bogus.
http://www.wired.com/news/culture/0,...l?tw=rss.index

Joel Garry wrote:
> Kind of a reach, but are you tracing?
> https://metalink.oracle.com/metalink...,1,1,helvetica
>
> Just for jollies, check MAX_TRANSACTION_BRANCHES and
> DISTRIBUTED_TRANSACTIONS on the 8 side (though it looks like the 10
> side is throwing the error - are you sure it doesn't have more links?).
>
> I think you are going to have to have a 9 in between.
>
> jg
> --
> @home.com is bogus.
> http://www.wired.com/news/culture/0,...l?tw=rss.index

Thanks for the reply, Joel.

Yep, I tried tracing, and also messed with DISTRIBUTED_TRANSACTIONS on
the 8 side. Tried bumping it up to 80, also tried reducing it to zero.
No impact either way. I knew this was a long shot, but I always work
under the assumption that "there are unknown unknowns" so hate to give
up without checking around.

Well, I'm not sure what happened, but when I tried to demonstrate the
failure to a hired gun this morning, there was no failure, everything
worked!