Communications Server for Data Center Deployment

There are several examples provided in the Enterprise Extender Implementation Guide (Redbook SG247359.pdf). You can get this from the Comm. Server Educational Material Website:

Education Materials for Communications Server for Data Center Deployment

  Chapter 4 has VTAM configuration examples, then Chapters 6 and 7 have AIX, Linux (respectively) with VTAM Switch Major Node examples.

Jeff L Smith

Comm. Server Development

Jeff,

Thanks for the references.  I will review and let you know if I have any further questions.

Regards,

Wayne Murray

Jeff,

I reviewed the documentation you referenced for defining the CS on Linux SNA node to VTAM via EE and it seems to agree with how we have it defined.

We are able to activate the node on VTAM and it connects and its VTAM PU comes active.  Status from the Linux side of the node also indicates that it is connected and active.

On the Linux side there is an LU defined for a CICS emulation region that is not running yet.  This LU defined under the active SNA node on the Linux side shows as active on Linux.  Yet when we display their PU on the mainframe this LU does not show up as I expected.  Is there something we have to do on VTAM to define this LU under their PU?  I expected it to be dynamic.  I suspect there is something not setup properly on the Linux side. 

Thanks,

Wayne Murray

If the PU is active on VTAM and the Comm. Server, then there is a DLUS/DLUS connection between the CS Linux and VTAM that will manage the PU/LU activity.

The 2 systems only use the "pu_id" (CS Linux) and "IDBLK/NUM" (VTAM) to match up the same PU each side has configured. The LUs are reference between each system using only the LU number (CS Linux) or LOCADDR (VTAM). If these match up, then they both will be reference the same LUs. (Section 7.3.3. of the Enterprise Extender Guide) 
Could the LU number be different between the 2 systems, like 1 on Linux and 2 on VTAM? 

The /var/opt/ibm/sna/sna.err (Error log file) on Linux will show any errors in activition. If you have Auditing activated, the sna.aud file in the same directory will show the PU and LU being activated for SSCP and for LU-LU sessions. To activate auditing on Linux, the command is "snaadmin set_log_type,audit=YES". 

The "snaadmin status_dependent_lu" will show the PU name and LU name on Linux where you can match up numbers against the LOCADDR configured on VTAM.

Jeff L Smith

Comm. Server Development. 

Additionally, If the LUs being used for CICS are LU6.2 (independent LUs), then VTAM would not need to define any LUs on the Comm. Server. The only VTAM configuration needed is a PU definition, which is the CP Name of the Comm. Server on Linux. CICS can use both dependent and independent LUs. The Enterprise Extender Guide sections 7.3.3 and 7.3.4 cover these for using EE.

Jeff L Smith

Jeff,

In this case this LU under the CS LInux PU will be an LU 6.2 independent LU.  I have verified that our VTAM has start option DYNLU=YES.  So, as I understand it, I do not have to define this LU under the PU on VTAM.  It should be defined dynamically when the CS for LInux PU connects to VTAM, correct?  And should it then show up under the Linux PU when we display it on VTAM?

Thanks,

Wayne Murray   

It may show up under the VTAM PU defined for the Linux if the CICS LU is on the same VTAM.

It may not show up if CICS is on another SNA node. In this case, the EE RTP connection may pass through the Network Node VTAM using ANR. It would show up on the CICS node dynamically.

LUs that are defined on the CS Linux will only show up on VTAM if the sessions the are ending on the local VTAM. Any LU the uses the Network Node VTAM to have its session routed to another VTAM will not show up on that Network Node. It will show up on the end point of the session. 

You may be looking for the LU, but the LU is not ending it session on the VTAM, just passing through on it link into the network. If the CICS LU is local, then it will show up. You should see it active with the Mode table it is using. 

Jeff L Smith

Jeff,

We have the Linux PU connected and the LU shows up as you indicated.  

So, now we are trying to establish a connection between CICS on the mainframe with the CICS emulator on the Linux side represented by that LU 6.2 on the Linux system.

Issue is that when we define the LU on Linux and activate it, it tries to establish a connection with the mainframe CICS system even before the emulator CICS region on the Linux side is started.  CICS on the mainframe receives a CNOS with the proper number of sessions and winners/losers, (8,4), but only 5 of the 8 sessions come active and they are all secondary.  Seems odd to me that there is any connection at all before the CICS emulator on Linux is even started.

So, it is obvious to me that something is defined wrong on the Linux side that we have to sort out.

But I want to thank you for your help getting us this far. 

Regards,

Wayne Murray

Thank you, Wayne, for the status on the LU6.2 setup between Linux and the VTAM CICS system.

 Coming from a VTAM background, it is probably normal to define a SNA resource then activate the definition. This does not actually activate any sessions. On the distributed side (Linux, Windows, AIX, other platforms), defining the SNA LU performs those operations together. If you "activate" the LU on Linux, you provide a partner LU and mode to use. This will start a set of sessions (defined by the mode table parameters) even if there are no APPC applications ready to use them. 

Some CICS admins do this so that the CICS session activation does not spend time waiting on the sessions to start. They sessions are there and available when CICS starts. Sort of like having a leased phone connection and being able to send data  immediately with the other side listening and not having to wait for the overhead of setting up the call.

The mis-match of expected sessions with regard to what actually get started may be due to how the mode table is defined on both sides. The VTAM side may have 8,4 for total sessions and owner sessions, but the Linux side my only have 8,2. You may need to check on the number of "conwinners/conlosers" in the Linux config to see that they have 4/4 for the total of 8 sessions.

Jeff L Smith

Jeff,

With regards to the issue where CICS was not starting or binding all of its session winners when the connection to the Linux CICS emulator was started.  I discovered that on the SESSIONS definition in CICS on z/OS, the AUTOCONNECT parm was set to NO.  Based on this CICS was not starting its session winners when the connection came active.  Instead, it starts them as needed when application programs allocate a session.  I have set AUTOCONNECT to YES and now when the connection is started all 4 of our winner sessions get started.  So, we have a total of 8 active sessions, our 4 winner sessions, and the 4 winner sessions from the remote CICS region.  

But only 1 of the two CICS management sessions, the winner from our side, gets started.  The other one, the winner management session for the remote CICS region, is not started.  But when the connection is started from the other side both management sessions get started.

So, at least we get all of the application sessions started.  So, one final little issue to resolve with the management session.

Regards,

Wayne

Jeff,

Thanks for your input on the management sessions.  I will restrain my anxiety over having just one management session active.

On another note.  I have a licensing question for CS for Linux.  If we were to use Linux on z/OS and we needed multiple Linux/zOS images with CS on them, all on one z/OS mainframe, would we still need a separate license for each Linux/zOS image with CS on it?  Or just one since they are all on the same z/OS machine?

Regards,

Wayne Murray

The Communications Server for Data Center Deployment (CSDCD)  is licensed 2 ways, Concurrent User, and PVU (Per Processor, or Per Core). The price point is about 500 Concurrent Users. When you need more than this, then the PVU licensing would be a better deal.

The CSDCD Concurrent User license does not consider the number of servers. So, you can have a System Z VM LPAR with a number of Linux images running. The Concurrent User license would cover all the users in all the images. The CSDCD server logs the Concurrent Users in the /var/opt/ibm/sna directory. Refer the Diagnostic Guide (czx01302.pdf) at 

https://www.ibm.com/support/pages/communications-server-data-center-deployment-linux-library in the section on License Management Tracking and the SNA usage log file.

Jeff L Smith