FIX: Network Error During RPC May Hang New Connections
  
PSS ID Number: Q100958
Article last modified on 10-31-1994
 
4.20
 
OS/2
 

----------------------------------------------------------------------
The information in this article applies to:
 
  - Microsoft SQL Server version 4.2 for OS/2
----------------------------------------------------------------------
 
BUG# OS/2: 1695 (4.2)
 
SYMPTOMS
========
 
If a serious network error occurs while a site handler is active, and
a Remote Procedure Call (RPC) is attempted across that site handler,
future connections to the originating SQL Server may stop responding
or hang. Current connections are unaffected. On some versions of SQL
Server, one additional connection can be made before the new
connections stop responding, and on others the problem starts with the
next connection.
 
This problem can be replicated by disconnecting the network between
two machines that have active site handlers. Initiate another RPC
(such as server2...sp_who) and wait for the error to be reported. Now
log in to the SQL Server that originated the RPC. If you are able to
log in, log out and try again. Existing connections should be OK,
however, every attempt to log in to the SQL Server should not respond.
 
WORKAROUND
==========
 
If you are experiencing this problem, try setting the timeout
parameter to False. This causes the site handler to quit after one
minute of inactivity. This minimizes the amount of time the site
handler is active. Because the problem only happens when a network
error occurs, this may minimize the occurrence.
 
STATUS
======
 
Microsoft has confirmed this to be a problem in SQL Server versions
4.2 and 4.2a for OS/2. This problem was corrected in SQL Server version
4.2aK10. For more information, contact your primary support provider.
 
Additional reference words: 4.20 4.2a 4.20a
KBCategory: kbother kbtshoot
KBSubcategory: SSrvServer
 
=============================================================================
 
Copyright Microsoft Corporation 1994.
