[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference pamsrc::objectbroker

Title:ObjectBroker - BEA Systems' CORBA
Notice:See note 3 for kits; note 5 for training; note 1134 for releases
Moderator:TLE::PARODId
Created:Tue Jul 11 1989
Last Modified:Thu Jun 05 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:1413
Total number of notes:6391

1395.0. "%OBB-E-INV_NOTAUTHORIS" by KERNEL::SIMPSONR (fred) Wed Apr 23 1997 08:47

VMS V6.1    DECedi V2.1c

I have a customer that has a problem with stuck documents that is currently
being worked on by engineering. To workaround the stuck documents they have 
written a routine that goes through a track list (list of available docs) and
makes sure that all documents can be fetched. The routines they have written
are all in Cobol and a 'C' interface has been written so that they can
communicate with DECedi V2.1c. Sometimes they receive the following errors in
the DECedi Error log file:

 DECEDI-I-ORBSTATMSG, Object Broker satus message =
 %OBB-E-INV_NOTAUTHORIS, client user is  not authorised to access server
 DECEDI-E-ORBERROR, An error was encountered calling the Object Broker routine
 CORBA_BOA_activate_obj,
 decedi condition logged

 DECEDI-I-ORBSTATMSG,
 %OBB-E-INV_SRVNOTREG, server is not registered.
 DECEDI-E-ORBERROR, An error was encountered calling the Object Broker routine
 OBB_BOA_DISPATCH
 decedi condition logged

 DECEDI-I-ORBSTATMSG,
 %OBB-E-INV_SRVNOTREG, server is not registered.
 DECEDI-E-ORBERROR, An error was encountered calling the Object Broker routine
 OBB_BOA_IMPL_UNREGISTER

This then kills their API programs. They then stop DECedi. All the OBB_pid
processes do not close down and so have to be deleted by hand. They then start
DECedi up again and it all works again.

This problem appears to occur when they have 60/70 docs being stuck at any one
time. Could it be that the max no of server instances has been reached? He
says that at any one time 2 or 4 copies of the server could be be going through
their tests and hammering Object Broker. 

Also on separate occasions they see the following problem:

 %DECEDI-I-ORBSTATMSG, Object Broker status message=
 %OBB-E-COM_BADUUID, Bad UUID '<NIL>' failure OBB dispatch
 OBB_BOA_DISPATCH
 decedi condition logged

Any ideas?
Richard Simpson.
T.RTitleUserPersonal
Name
DateLines
1395.1KERNEL::SIMPSONRfredWed Apr 23 1997 08:4851
The problem is still with them even though the stuck documents problem has
been resolved. I rang the customer and he said that the problem is still
occurring. They get an extra error that now mentions deactivate instead of
activate:

 DECEDI-I-ORBSTATMSG, Object Broker status message =
 %OBB-E-INV_NOTAUTHORIS, client user is  not authorised to access server
 DECEDI-E-ORBERROR, An error was encountered calling the Object Broker routine
 CORBA_BOA_deactivate_obj,
 decedi condition logged

Then he gets the following errors logged after the DECEDI$ERROR log errors in
a file called DECEDI.ERR. This file is found in SYS$LOGIN of the detached
server that is using Objectbroker.

 Module : DECEDI_API_FETCH , %OBB-E-BOA_OBJNOTACTIV, The Object Reference is Not
  Active.

 Module : DECEDI_API_FETCH , %ACAS-E-INV_HOSTNOTFND, A host node could not be
  found to execute the request.
 -ACAS-E-INV_NOTAUTHORIZ, CLient user is not authorized to access server.

 Module : DECEDI_API_TRACK , %ACAS-E-INV_HOSTNOTFND, A host node could not be
  found to execute the request.
 -ACAS-E-INV_NOTAUTHORIZ, CLient user is not authorized to access server.

 Module : DECEDI_API_FETCH , %ACAS-E-INV_HOSTNOTFND, A host node could not be
  found to execute the request.
 -ACAS-E-INV_NOTAUTHORIZ, CLient user is not authorized to access server.

 Module : DECEDI_API_TRACK , %ACAS-E-INV_HOSTNOTFND, A host node could not be
  found to execute the request.
 -ACAS-F-COM_INSUFVM, Insufficient Memory

 Module : DECEDI_API_TRACK , %ACAS-E-INV_HOSTNOTFND, A host node could not be
  found to execute the request.
 -ACAS-E-INV_DISCONN, The connection has been disconnected.
 -ACAS-E-INV_TRANSERR, Transport error 'broken pipe'.

 Module : DECEDI_API_TRACK , %ACAS-E-INV_HOSTNOTFND, A host node could not be
  found to execute the request.

 Module : DECEDI_API_FETCH , %ACAS-E-INV_HOSTNOTFND, A host node could not be
  found to execute the request.
 -ACAS-F-COM_INSUFVM, Insufficient Memory

Please can someone have a look at this problem and provide some ideas as to what
information is required to further this problem,

Cheers,
Richard.
1395.2METSYS::THOMPSONWed Apr 23 1997 17:077
One thing we have observed here is that Objectbroker Agent is is exhausting
it's Pagefile quota.

Any ideas what can cause the agent to leak so?

M
1395.3What version of ObjectBroker?RECV::VLATASWARNING: Do not swallow the battery doorThu Apr 24 1997 10:5615
    
    What version of ObjectBroker is being used?
    
    There was a problem that was corrected for V2.6 where failed proxy
    lookups would cause a >1K per-invoke memory leak. The output in
    1395.1 indicates that there are proxy lookup failures so this could
    be why you are running out of memory.
    
    This was QAR 2872. Note that invokes can appear to be working OK and
    you will still get this leak. For example: if you don't have a proxy
    setup for a specific host/user combination but do have one setup for
    a host/default combination, it will leak when the check for a specific
    host/user proxy is made and then it will match the host/default proxy.
    
    Tony
1395.4METSYS::THOMPSONThu Apr 24 1997 15:0828
Hi,

This is ObjectBroker V2.5A. OpenVMS VAX V6.x, TCP/IP transport. 

What we did was alter obb$startup.com to give the agent about 200,000 block
Page File quota. 

Their test has about 5000 operations (fetches).
When they ran the test   they got a 'leak' of 30,000 blocks.
They did a re-run of the same test and got a leak of 'only' 9000 blocks.

So we passed their test but now that they are able to estimate the
memory leak they realise they would have to re-start the Agent every
day. This is not really acceptable, they estimate 10 days as a reasonable
memory leak crash period.

They are now using CORBA interfaces only.
The DEC/EDI client and server processes do not appear to be leaking, only
the agent.

The Authorization problems are now resolved. So it does look as though
those failures were caused by lack of memory.

I don't know why the Agent leaks so badly for this customer. 
We can't upgrade to V2.7 as this release has problems when we register
our interfaces.

Mark
1395.5METSYS::THOMPSONThu Apr 24 1997 15:116
One last detail ... this is a single node system. All operations are
launched from a single account. So I don't know why it's trying to
authorize anyway. The proxy database is empty.

Mark