RUNUBE / RUNUBEXML on a Windows Cluster

Has anyone ever used RUNUBE/RUNUBEXML with a Windows Clustered Enterprise Server?

We can run it fine on our DEV/TEST Enterprise Server so I know we are using the proper format but it will not work on the production clustered environment.

I'm fairly certain that the problem is being caused because actual physical server names are ERP1 and ERP2 with a Windows cluster name of ERPCLUST but as when you define a cluster in JDE you give a third name like ERPSVR.

We we run the RUNUBE command a log is created for the UBE being submitted but after a while IPC errors appear in the log indicating that it cannot communicate with the METADATA kernel. (see errors below)The job also never appears in the WSJ table for the cluser name becuase I believe the RUNUBE command is not cluster aware.

I can try changeing the ipcstart value but I don't really think that will help.

10508/18764 MAIN_THREAD Mon Oct 20 12:43:59.908000 ipcmisc.c299
process 10508 <RUNUBE> registered in entry 49

10508/18764 MAIN_THREAD Mon Oct 20 12:43:59.923000 Runube.c724
Startup for User=BVRUNNER, Env=PD812, Role=*ALL, Report=ROOO6P, Version=XJDE0001

10508/18764 MAIN_THREAD Mon Oct 20 12:43:59.923001 Runube.c751
bAPListening=0

10508/18764 MAIN_THREAD Mon Oct 20 12:45:30.004000 SpecCnvrt.c2002
METADATA - Attempted to send Metadata Kernel a message 3 times and did not recieve a reply. Returning failure to the caller. The Metadata Kernel is Kernel #30 in the jde.ini and the message ID being sent to the kernel is 15751.

[ QUOTE ]
Has anyone ever used RUNUBE/RUNUBEXML with a Windows Clustered Enterprise Server?

We can run it fine on our DEV/TEST Enterprise Server so I know we are using the proper format but it will not work on the production clustered environment.

I'm fairly certain that the problem is being caused because actual physical server names are ERP1 and ERP2 with a Windows cluster name of ERPCLUST but as when you define a cluster in JDE you give a third name like ERPSVR.


[/ QUOTE ]

Mike,

You will need to use the MS Cluster name. In your above example, use ERPCLUST. JDE is not a cluster aware application, so the physical names are meaningless. But your application will need to refer to the Windows cluster name for running JDE, and subsequently for running UBEs either through the application or through runUBE.

If your database in MS SQL, you will need yet another alias, since SQL does not want to use the Windows cluster name. In my case, I need to keep track of four different names for the production portion of my enterprise server.

Physical box 1 - xxJDEDB70
Physical box 2 - xxJDEDB71
MS Cluster Name - xxJDEDB82
SQL Cluster Name - xxJDEDB83

The good news is that it works. We have been running Tidal Scheduler (which uses the runUBE command to execute jobs) for years now and it works just fine on a cluster.

Gregg

Gregg,
I'm a little unclear on your answer. What is the name of the Server in JDE that points to your MS Cluster? When we installed 8.12 in JDE we had to define a separate Logical Server name that was different that the MS Cluster name.
In other words:
Physical Machine: ERP1
Physical Machine: ERP2
MS Cluster name: ERPCLUST
JDE Logical Server: ERPSVR

Tell me if these are the steps you use to test RUNUBE:
1. Log on to xxJDEDB82 using the service user that is used to start the JDE services(I assume you are using the admin tool Remote Desktops becuase RUNUBE will not run with just RDP)
2. Type in your command like:
RUNUBE JDE JDE PD812 *ALL R0006P XJDE0001 QBATCH BATCH HOLD SAVE

Now when you log into JDE and go to WSJ "What server do you see your job running under?



(BTW we have another cluster for just the SQL DB called ERPDB with nodes ERPDB1 and ERPDB2)
Physical Machine: ERPDB1
Physical Machine: ERPDB2
SQL Cluster: ERPDB
JDE DB Datasource: ERPDB

Mike

A few differences here - our prod environment is XE. We're working on upgrading to 8.12, but it's not there yet. When my Tidal agent runs a job, it runs the job from a dedicated fat client. That is where the runUBE command is being executed from. The actual ube is running on the clustered enterprise server. JDE refers to that enterprise server by the MS cluster name xxJDEDB82. That is the name that pops up in work with server. We haven't tried running Tidal in 8.12 or 9.0 yet so I don't know how those releases will deal with runUBE.

I do believe that clustering in 8.12 is really different than clustering in XE. I know that the errors I am getting are related to a kernel called METADATA and I know that XE doesn't even have that Kernel.

Thanks for the help anyway and good luck when you impliment the clustering in your 8.12 env. I will tell you it was really painful. I ended up on a conference call with about 10 JDE employees helping them re-write the clustering portion of the 8.12 install guide.

I'm still looking for any 8.12 users who have tried RUNUBE on a cluster. ANYONE???????

I solved my own problem.

I added the following line to the JDE.INI file on the cluster

[CLUSTER]
PrimaryNode=EONESVR

EONESVR is the JDE Cluster name I defined within JDE NOT the Windows cluster name.

mtrottier said:

I solved my own problem.

I added the following line to the JDE.INI file on the cluster

[CLUSTER]
PrimaryNode=EONESVR

EONESVR is the JDE Cluster name I defined within JDE NOT the Windows cluster name.

Click to expand...

Mike,

Hopefully you still frequent this board. I'm having what appears to be the exact same issue. However, I am by no means an expert or even user of JDE anything so bare with me. I've followed along for the change points to the JDE.ini file but I'd be most appreciative if you could help me. By "defined within JDE" what do you mean? This is entirely separate to any of the Windows names for the cluster resources, correct? If so, where would I find where the cluster name was defined in JDE?