APPC/CPIC App Unable to Allocate Conversation over Dependent LU (225207)


The information in this article applies to:

  • Microsoft SNA Server 3.0 SP3
  • Microsoft SNA Server 4.0 SP1
  • Microsoft SNA Server 4.0 SP2
This article was previously published under Q225207

SYMPTOMS

An APPC or CPI-C application may fail to get an active conversation with a remote APPC LU if a previous conversation with a different remote APPC LU using the same dependent local APPC LU has ended. For example, if there was a conversation over a dependent local LU, "LOCAL," and a remote LU, "REMOTE1," an APPC or CPI-C application may fail to get an active conversation over dependent local LU "LOCAL" and remote LU "REMOTE2" even if the previous conversation has ended.

When this occurs, the APPC or CPI-C application call to [MC_]ALLOCATE or CMALLC never completes. If the application does not use a timeout, the ALLOCATE will appear to hang since the call does not complete. If the application does have a timeout, the timeout will cancel the outstanding ALLOCATE, and return an error.

For additional information about a related problem in earlier versions of SNA Server, please see the following article in the Microsoft Knowledge Base:

180416 APPC or CPI-C TPs May Fail When Using Dependent LU 6.2 Sessions

NOTE: This problem can only occur when using Dependent LU 6.2 sessions and does not occur when using Independent LU 6.2 sessions.

CAUSE

A similar problem was corrected in earlier versions of SNA Server as described in article Q180416. SNA Server was modified to send an UNBIND Cleanup to the host to end the previous session that is no longer being used. In this case, the UNBIND Cleanup is treated as an error by the CICS Region on the host, so CICS attempts to re-establish the session over the local APPC LU/remote APPC LU that was previously being used. CICS sends a BIND to the SNA Server to re-establish this session. SNA Server responds with a BIND +RSP, which causes the ALLOCATE for the new session over the same local APPC LU and a different remote APPC LU to hang because the local APPC LU is already in use.

RESOLUTION

To resolve this problem, obtain the latest service pack for SNA Server version 4.0. For additional information, click the following article number to view the article in the Microsoft Knowledge Base:

215838 How to Obtain the Latest SNA Server Version 4.0 Service Pack


STATUS

Microsoft has confirmed that this is a problem in Microsoft SNA Server versions 3.0 SP3, 4.0 SP1, and 4.0 SP2. This problem was first corrected in SNA Server version 4.0 Service Pack 3.

MORE INFORMATION

After applying the update, SNA Server will issue an UNBIND Normal to end the previously used session instead of an UNBIND Cleanup. This is not treated as an error by CICS on the mainframe, so it will not attempt to re-establish the previous session.

This problem only occurs with dependent LU 6.2 LUs because they can only support one concurrent session at a time. Independent LU 6.2 LUs can support multiple parallel sessions concurrently, so a new session request to a different remote LU can be fulfilled without ending the previous session.

The following is a sequence of events that can lead to this problem assuming the following APPC LUs are defined in the SNA Server configuration:

Local APPC LU: LOCAL (defined as a dependent LU 6.2 LU)
Remote APPC LU: REMOTE1
Remote APPC LU: REMOTE2
APPC Mode: APPCMODE
  1. An APPC or CPI-C application successfully establishes a session over the following APPC triplet: LOCAL/REMOTE1/APPCMODE.
  2. The application executes a transaction over the previously established session and ends the conversation.
  3. An APPC or CPI-C application requests a session over the following APPC triplet: LOCAL/REMOTE2/APPCMODE.
  4. SNA Server issues an UNBIND Cleanup to end the original session established in Step 1.
  5. SNA Server sends an INITSELF command to the host to solicit a BIND for the new session over LOCAL/REMOTE2/APPCMODE.
  6. The host responds with an INITSELF +RSP for this new session request.
  7. The host then sends a BIND to re-establish the original session over LOCAL/REMOTE1/APPCMODE.
  8. At this point, the requested session over LOCAL/REMOTE2/APPCMODE cannot start until the original session is ended. This causes the ALLOCATE to be queued such that it appears to hang or timeout depending on the APPC/CPI-C application.


Modification Type:MinorLast Reviewed:9/22/2005
Keywords:kbHotfixServer kbQFE kbbug kbfix kbQFE kbsna400sp3fix KB225207
©2004 Microsoft Corporation. All rights reserved.