ORA-25401 – “Cannot continue fetches”

Tonight I had a customer call me requesting help.  When asked what the problem was, they did not really know.  All they knew was that production was having a problem and it needed to be fixed.  Customer suggested that I contact the on-site support; so after tracking down a phone number I was able to get in touch with the support hat was on-site.

After spending a few minutes talking with support, I had an idea of what the issue was.  I then ask if there were any associated error messages.  The message that was sent over was:  “ORA-25401: Cannot continue fetches”.  As most DBAs would be saying is: “Another random error message”.  Sure it is random, but that is what makes this interesting.

The customer is running a 2-node RAC system; where we (customer and myself) took down the second node in the RAC to work on another problem.  When the second node came down all the sessions failed over to the first node; however, the sessions didn’t connect as expected.  So, when the session tries to continue processing it receives an “ORA-25401: Cannot continue fetches” error.

How did we resolve this?  I had to disable TAF on the client side through the TNSNames.ora file. (items marked with an asterisk (*) were removed).

Before:

PRODDB =

(DESCRIPTION =

    (ADDRESS = (PROTOCOL = TCP)(HOST =hostname-vip)(PORT = 1521))

   * (ADDRESS = (PROTOCOL = TCP)(HOST = hostname-vip)(PORT = 1521))

    *(LOAD_BALANCE = yes)

    (CONNECT_DATA =

      (SERVER = DEDICATED)

      (SERVICE_NAME = ORASID)

      *(FAILOVER_MODE=(TYPE=select)(METHOD=basic)(RETRIES=120)(DELAY=5))

    )

  )

 After:

 PRODDB =

(DESCRIPTION =

    (ADDRESS = (PROTOCOL = TCP)(HOST = hostname-vip)(PORT = 1521))

    (CONNECT_DATA =

      (SERVER = DEDICATED)

      (SERVICE_NAME = ORASID)

    )

  )

Once the TNSNames.ora file was updated, the connections were reset and worked as expect from the applications view.  

Now, this is not the desired solution with a RAC environment since we are always trying to keep high-availability.  In some cases, it will due, especially on a two-node RAC configuration.

 Feel free to comment.  Always good to see how other solves similar problems.

 Thanks

Please follow and like: