| Author |
Message
|
| mverh |
 Posted: Thu Jul 17, 2003 12:32 pm Post subject: WBIMB - Trading known problems for unknown problems |
 |
|
Voyager
Joined: 06 Mar 2002
Posts: 97
|
I've installed WBIMB V5 and migrated some flows and message sets. Needless to say I have encountered some challenges. I 'll share them here and would like to hear from other's on their challenges.
I first installed vanilla WBIMB V5 (w/o Rules and Formatter extension).
My installation is on a W2K standalone box.
My layout has a queue manager for the ConfigMgr and a queue manager for my Broker.
The install went smooth.
Created ConfigMgr with no problems. Note there is no longer a database required for the MRM.
Created broker with no problems.
Installed a CVS server on another W2K image and managed to install the CVS perspective with no problems. Note sspi security is not supported. I am using pserver. I haven't been able to validate domain users on the CVS server via the eclipse workbench, but locally defined users on the CVS server work. I have a question in with IBM on how to specify a domain user when logging into CVS. I have no problem using sspi from WINCVS. This challenge appears to be an eclipse issue. In any case I can use CVS for version and concurrency control.
Imported message flows from a WMQIB V2.1 image. There were some issues with eSQL being incorrect. For example, Path is now a reserved keyword and we used it in some of our compute nodes as a variable name. Also, some flows had connections between nodes that were dropped. Note that the mqsimigratemsgflows utility is pretty dumb. For example, it creates a log file in the directory that you execute it from. If this directory is the same directory where your V2.1 msgflow is then the utility will try to also import the log! The nice thing is the eclipse tooling highlights your errors and makes it easy to fix problems. It creates a task list for you!
Imported some message sets that represent Cobol layouts. Believe it or not this went without any problems.
Built a broker archive (bar file). So now IBM build bars! The bar contains your message flows and message sets. The bar is then dragged onto you execution groups and deploys automatically. The concept of a bar is great since a problem I have encountered numerous times is deploying flows w/o deploying a dependant message set only to find out at run time. Now when you build your bar it is easy to see if you have left out flows or message sets. I am not convinced that the automatic deploy is a good thing though.
I am having challenges importing a cobol copybook. The instructions don't work. I can import the file containg the copybook per the instructions but when I try to build the Message Set Definition have been unsuccessful. I have a ticket open with IBM on this.
I applied CSD01 and appear to have lost options on the workbench.
Overall though I think the tooling is better and once you get used to differences between the Control Center and the eclipse tooling you'll be ok.
I have yet to look into the new features (http node, namespace support etc) I am simply trying to get my V2.1 environment running in a V5 environment.
Comments appreciated.
_________________
Marc Verhiel
IBM Canada Ltd. |
|
| Back to top |
|
 |
| jefflowrey |
| Posted: Fri Jul 18, 2003 5:17 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
<i>I applied CSD01 and appear to have lost options on the workbench. </i>
My understanding, particularly from reading the most recent readme, is that you should not have done ANYTHING with the product post-install until you have installed CSD01.
The sketchy availability of the Windows CSD01 is one of the reasons I am roughly two weeks behind where I wanted to be on setting up WMIMBv5. |
|
| Back to top |
|
|
| mverh |
| Posted: Fri Jul 18, 2003 5:37 am Post subject: |
|
|
Voyager
Joined: 06 Mar 2002
Posts: 97
|
After encountering the error I decided to review the readme.html and saw the comment indicating that you must not do anything before installing CSD01. So I have decided to uninstall everything and then reinstall. Hopefully everything I put in CVS will be easily recovered.
Note on the import of a cobol copybook problem. The file being imported must have an extension of .cbl to be seen when you select the cobol radio button on the new Message Definition File dialogue.
_________________
Marc Verhiel
IBM Canada Ltd. |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Wed Jul 23, 2003 11:59 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
I thought I would add my own experiences (so far) with v5 here, rather than starting my own thread.
My basic plan is to build a test platform, and use that to migrate the apps I currently have on my development/uat platform (two brokers, one machine). Then I will convert the uat platform, and promote my converted apps from the test platform. After successfull uat testing of the converted apps, the production server will get migrated and the apps repromoted from uat.
I did an initial install using the custom options, so as not to install a UserName server since I don't use topic level security. This seemed to cause some issues with what was installed or not, so I reinstalled using a Typical install and have not set up a UserName server.
Somewhere along the way, I had to manually create the mqbr* groups in the local security environment. They did not get created for me.
The Getting Started wizard is nice in that it will build everything for you, but it has some drawbacks. Particularly for setting up an environment that needs to be at least somewhat aware of Windows Domain security, I do not recommend using the Getting started Wizard. Just use the mqsicreate* commands like normal.
One big issue I had with the Getting Started wizard is that it would not let me specify a Domain level user for the service user ID. It would only let me specify a local user. Another issue I had was that the user running the Toolkit had to have sufficient rights to the necessary DB2 instances for the Getting Started wizard to be able to create the databases for you - this is another reason for using the command line options.
I had to play around a bit with the security domain awareness of the config manager to get it to work with my local setup requirements (authenticate domain level users, but use the local groups).
I was able to create a simple message flow with the same ease as before, but it took me a little bit to get used to the new deployment mechanisms and concepts. You can't deploy a message flow anymore, only a broker archive file. I'm still also trying to get used to referring to the ConfigMgr as the "Domain" in the new tooling.
As I work with this more, I'll post more comments. |
|
| Back to top |
|
|
| JLRowe |
| Posted: Wed Jul 23, 2003 2:01 pm Post subject: |
|
|
Yatiri
Joined: 25 May 2002
Posts: 664
Location: South East London
|
My biggest gripe so far is that you cannot refresh a BAR, the documentation says you have to remove resources and add them back again!
I know you can use the 'Run on server' option for a quick deploy, but it would be nice to refresh a BAR.
Also, the original CSD01 was pulled sometime last week and it now has an updated fix.
I just debugged some ESQL - you can now step through the contents of a compute node - excellent. |
|
| Back to top |
|
|
| kirani |
| Posted: Wed Jul 23, 2003 2:11 pm Post subject: |
|
|
Jedi Knight
Joined: 05 Sep 2001
Posts: 3779
Location: Torrance, CA, USA
|
Did it automatically install IBM Agent Controller on your box? For me it didn't, and I cannot find it on CD either! Finally, I installed it from my preview version.
How can i check the version of the Agent Controller installed on my box?
_________________
Kiran
IBM Cert. Solution Designer & System Administrator - WBIMB V5
IBM Cert. Solutions Expert - WMQI
IBM Cert. Specialist - WMQI, MQSeries
IBM Cert. Developer - MQSeries
|
|
| Back to top |
|
|
| benc |
| Posted: Wed Jul 23, 2003 11:45 pm Post subject: |
|
|
Newbie
Joined: 30 Jul 2002
Posts: 3
|
After digging throught the stack of CDs which have similar names....
The Agent Controller is on another CD (Supplementary) which has to be installed separately. One way I know to find out the version is to check the
Windows registry and look for the folder:
HKEY_LOCAL_MACHINE\SOFTWARE\IBM\IBM Agent Controller
ben
---
[quote="kirani"]Did it automatically install [b]IBM Agent Controller[/b] on your box? For me it didn't, and I cannot find it on CD either! Finally, I installed it from my preview version.
How can i check the version of the Agent Controller installed on my box?[/quote] |
|
| Back to top |
|
|
| mverh |
| Posted: Thu Jul 24, 2003 9:53 am Post subject: |
|
|
Voyager
Joined: 06 Mar 2002
Posts: 97
|
Some more updates.
I agree with Jeff that the Getting Started wizard is not too useful. I created my ConfigMgr and Broker via the command line. Since I already had V2.1 installed I didn't require W2K groups to be added so I can't say if I encountered that particular problem.
I am also trying to figure the best way to migrate from sysprog test->dev->test->QA->Prod. I am in the process of migrating all my current V2.1 flows and messages sets and fixing what doesn't work prior to exploring new options. One thing we are looking at is running the V2.1 Control Center and the Eclispe Workbench (EW) on the same desktop since this will be a neccessity when you run V2.1 and V5 in parrallel. I will update this thread once we figure it out.
Some things I've come across.
IBM says CVSNT V2.0.5 is not supported by Eclipse. Here's a link they gave me as a reference.
https://bugs.eclipse.org/bugs/show_bug.cgi?id=35961
Now bearing this in mind I have had no problems interfacing my EW with CVSNT. I have added projects, updated them outside of EW and synchronized these changes back into the EW with no issues. The only issue I have is I can't use sspi and when using pserver the userid on the CVSNT box must be locally defined. So I am comfortable using CVSNT 2.0.5 as my CVS server for now.
The EW has some drawbacks compared to the CC. In the CC you could cut and paste between message sets I haven't found a way to do this in EW. Also in CC you had the ability to add a message set to a compute node and then drag elements from the message definition into the esql pane. This would help you code esql and minimize fat finger typos. I haven't discovered an equivalent in the EW.
When importing message flows we have noticed that opening a compute node's esql sometimes results in a duplicate "CREATE COMPUTE MODULE" code block to be generated. Simply deleting this code resolves the problem. Also, migrated compute node esql does not include CopyMessageHeaders and CopyEntireMessage procedures. This is not a problem but means your v2.1 esql will not be consistent with v5 esql.
I have also encountered problems where I get numerous "Invalid reference" tasks in my task list. I have discovered that simply checking a message set reference in the message flow preferences->references dialogue fixes this. It appears as simply changing the state of the reference i.e. checked->unchecked or unchecked->checked causes the tasks to be removed from the task list.
Adding a compute node to a flow causes some standard template esql to be added to the flow esql file. However, if I delete the compute node the esql is not deleted.
I've had a copy of instances where the javaw.exe representing the EW goes 100% cpu when I save changes. Killing the process and restarting seems to fix it. Based on this I recommend saving changes often.
HTH.
_________________
Marc Verhiel
IBM Canada Ltd. |
|
| Back to top |
|
|
| kirani |
| Posted: Thu Jul 24, 2003 10:10 am Post subject: |
|
|
Jedi Knight
Joined: 05 Sep 2001
Posts: 3779
Location: Torrance, CA, USA
|
I wonder if anyone was able to migrate plug-in nodes to V5. By migrate I mean creating/installing them in WMQI studio.
_________________
Kiran
IBM Cert. Solution Designer & System Administrator - WBIMB V5
IBM Cert. Solutions Expert - WMQI
IBM Cert. Specialist - WMQI, MQSeries
IBM Cert. Developer - MQSeries
|
|
| Back to top |
|
|
| mverh |
| Posted: Thu Jul 24, 2003 12:12 pm Post subject: |
|
|
Voyager
Joined: 06 Mar 2002
Posts: 97
|
Here's how to run Control Center V2.1 on the same desktop as Eclispe Workbench V5.
Zip up the directory containing your V2.1 CC. Copy the zip to the box where you want the CC to run. In my case this was the desktop that had the EW V5 installled. Unzip it. Update the mqsilcc.bat in the ..\Tool directory and add a PATH statement. The Path should include the CC V2.1 bin directory.
Example:
set PATH=C:\PROGRA~1\IBM\WEBSPH~1.1\bin;%PATH%
Note I used the 8.3 format for my directory names (less typing).
In a command window cd to the directory containing mqsilcc.bat. Invoke mqsilcc.bat.
That's it. Note I didn't try installing the V2.1 CC from the CD. Who knows maybe that'll work too.
HTH.
_________________
Marc Verhiel
IBM Canada Ltd. |
|
| Back to top |
|
|
| vmcgloin |
| Posted: Thu Jul 24, 2003 12:54 pm Post subject: |
|
|
Knight
Joined: 04 Apr 2002
Posts: 560
Location: Scotland
|
| Quote: |
| One thing we are looking at is running the V2.1 Control Center and the Eclispe Workbench (EW) on the same desktop since this will be a neccessity when you run V2.1 and V5 in parrallel. |
Sorry for butting in but I understood that the V5 tooling would be compatible with 2.1 brokers. Is this wrong? Is it possible to deploy to 2.1 broker from the v5 workbench??? I was hoping to just upgrade the windows side of our setup in the short term.
Thanks,
Vicky |
|
| Back to top |
|
|
| mverh |
| Posted: Thu Jul 24, 2003 1:59 pm Post subject: |
|
|
Voyager
Joined: 06 Mar 2002
Posts: 97
|
You can deploy from a V5 workbench/ConfigMgr to V2.1 brokers provided all the prereqs are met. However, I you can't connect a V5 workbench to V2.1 ConfigMgr. So until all your ConfigMgrs are at V5 you'll need a V2.1 CC.
On the topic of deploying from a V5 workbench/ConfigMgr to a V2.1 Broker. I don't think I am going to go down that road. When we upgrade our brokers I think we will delete them, upgrade the code, recreate the broker and deploy. I have had some bad experiences when a broker that is already in one ConfigMgr is attempted to be brought into another ConfigMgr. To get it to work we had to recreate the broker and then bring it into the new ConfigMgr.
_________________
Marc Verhiel
IBM Canada Ltd. |
|
| Back to top |
|
|
| kirani |
| Posted: Thu Jul 24, 2003 2:01 pm Post subject: |
|
|
Jedi Knight
Joined: 05 Sep 2001
Posts: 3779
Location: Torrance, CA, USA
|
Per online help, i should find sample code for plug-in's into install_dir\sample\extensions\nodes directory. But I am missing following files,
TransformNode.java
SwitchNode.lil
SwitchNode.java
SwitchNode.jar
SampleNodesProject.zip
_________________
Kiran
IBM Cert. Solution Designer & System Administrator - WBIMB V5
IBM Cert. Solutions Expert - WMQI
IBM Cert. Specialist - WMQI, MQSeries
IBM Cert. Developer - MQSeries
|
|
| Back to top |
|
|
| pfaulkner |
| Posted: Tue Jun 15, 2004 12:34 pm Post subject: |
|
|
Master
Joined: 18 Mar 2002
Posts: 241
Location: Colorado, USA
|
I just installed V1 ConfigMgr and Toolkit on a Win 2000 box. I added an existing v2.1 broker to the topology and then an EX Group. I was able to deploy the Topology without any problems.
I then migrated one of my flows and msgsets from a 2.1 Config Mgr, also without any issues.
When I came to deploying I got these errors:
The correlation name 'CREATE' is not valid....
In first 2 lines of the esql it complains about, SI has added these lines
CREATE COMPUTE MODULE "OB_RECEIVE_MANAGER_REQUEST_Check if Oneway message"
CREATE FUNCTION main() RETURNS BOOLEAN BEGIN
I assume the CREATE COMPUTE MODULE line is not supported on the 2.1 broker and hence the error???
But if this is the case, how do I deploy to a 2.1 broker? |
|
| Back to top |
|
|
| mverh |
| Posted: Tue Jun 15, 2004 1:05 pm Post subject: |
|
|
Voyager
Joined: 06 Mar 2002
Posts: 97
|
Don't know if this will help in your situation but try it and let us know:
Preferences->ESQL and Mapping->Code Generation
Select 2.1
_________________
Marc Verhiel
IBM Canada Ltd. |
|
| Back to top |
|
|
|
|
