Document ID:        45306.1
Subject:            Common Errors in Oracle Intelligent Agent
Last Modified:      16 Apr 97 
Author:             ASHRIVES


Common Errors
~~~~~~~~~~~~~

Once a snmp.ora is created the listener seems to hang when started.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Edit your snmp.ora and ensure you have your listeners name (the default
is 'listener') in the visible services.

You will also need a SNMP.INDEX.<listener name> line. 
 

You can't stop the agent
~~~~~~~~~~~~~~~~~~~~~~~~
While attempting to shutdown the Agent using '$ lsnrctl dbsnmp_stop', if
this action is aborted by a ctrl-Y, a group logical AGENT_STOP remains
defined causing the Agent to quit every time it is started up. To resolve
this, manually de-assign this group logical and then start the Agent, using
the command  '$ deassign/group/user AGENT_STOP'.

The agent won't start
~~~~~~~~~~~~~~~~~~~~~
If you do an  $ lsnrctl dbsnmp_start followed by an
              $ lsnrctl dbsnmp_stop 

and it shows the agent is not running then add the following lines
to the ora_agent:AGENT_START.COM file:

   $ define sys$error DISK$SUPPORT1:[ORACLE73.V7323.NETWORK.AGENT]error.log
   $ define sys$output DISK$SUPPORT1:[ORACLE73.V7323.NETWORK.AGENT]error.log
   $ set verify

Obviously change the directory specifications in the above to match your
system. When you have started the agent, check the error logfile produced for
errors.

Enterprise Manager shows job completed AND Failed on startup/shutdown 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 
When a database shutdown job is submitted from the OEM Console to the 
Agent running on VMS, the job completes correctly, and shows a "Job 
completed successfully" in the Job History details on the Console, but 
at the same time there is also another entry in the list that shows this 
job as failed. Bringing up the details of this failure on the Console 
shows "Output to job <nn> lost"". 
Checking on the VMS node where the Agent is running does indicate that the
database has been shutdown correctly by the Agent. Similar behavior is also
observed with the database startup job.

This is Oracle Bug 399692

Jobs are submitted, but never get scheduled or run
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This is the hardest one. Normally it indicates a problem in the agent.

Our recommendation is to try the following:

1. Stop the agent and the oracle listener.

2. Start the listener and the agent.

   Starting the agent should be instantaneous, that is, you should get the
   VMS prompt almost immediately back after typing lsnrctl dbsnmp_start.

   If you don't, try the following: 

    (a) At the VMS prompt do a $ show dev /files
    (b) Find any files that have agent in the filename or have a .q file
        extension. 
    (c) Kill the processes that have these files locked with the stop /id
        VMS command.
    (d) Try restarting the agent.

When you start the agent for the very first time, it creates the .q files 
in the ora_agent directory. It has been noticed in this situation that the
jobs from the OEM do not get reported back correctly from the Agent.
Simply shutting down and restarting the agent resolves this problem.
Subsequent shutdown and restart do not cause this problem either.
      
You can't use Tablespace Manager to Defrag a table
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
When you queue a job to defragent a table from tablespace manager, the job
fails. The output from the command is:

  successfully logged onto database
  successfully generated export file
  successfully generated temp file
  successfully locked reference tables
  successfully dropped fragmented table
  successfully imported export file
  succeerential constraints
  error running command: ambiguous command verb - supply more characters

This has been logged as Oracle bug number 478931. It will be fixed in the
next release of Oracle Agents under OpenVMS. (in the 7.3.3 release)