SAP PI on MSCS

[Former Member]

All

I am also facing problem in MSCS enviorment while performing dialog installtion for system copy(last step in Installation of target system in mscs mode) .

Below is the enviorment:

Homogenous system copy (SAP PI 7.0) from win 2003 32 bit x64 (standalone with MSSQL 2000) to win 2003 64 bit x64 (MSCS with MSSQL 2005)

Source system export completed

MSCS installtion done (by OS team)

MSSQL 2005 software successfully clustered with failover

ASCS installtion successfully done (instnace number 11)

SCS installtion successfully done (instnace number 61)

Configure first MSCS node installtion successfully done

Install Database Instance successfully done

Configure additional MSCS Node successfully done

Installed Enque replication server on both nodes (instnace number 21)

Installed Central instance on first MSCS node (instnace number 01)

Installing DI-Dialog Instance on add. MSCS node (error in the last step "start dialog instance")

Problem 1

The dialog instance doesnt started shows status ABAP active, J2EE status unknown (elapsed time 21:30) giving up.

When i check the SAPMMC, the dispatcher of Dialog instance is showing yellow J2EE status info unavailable....

Problem 2

For troubleshooting the above problem, i tried looging into the ABAP instance.... but the SAP instance doesnt connects through the virtual groups. It successfully connects throught the physical hostname of the server.... which is wrong. in MSCS enviorment end users will connect to the SAP instance via the virtual name (ASCS instance number).

I am able to successfully telnet to that ASCS virtual hostname port 3611 & 3911 but cant connect through SAPGUI

Waiting for a reply asap.

[Former Member · Accepted solution]

Abb,

The way MSCS works in NW2004s systems is different from earlier releases. With NW2004s, SAP introduced the concept of ASCS for high available systems. Since ASCS consists of Message server, Enqueue server, these are pretty much your SPOF (single point of failure). You only need to make sure that ASCS is available at all times. You then have multiple application servers (including CI) connecting to your message server.

Therefore you install ASCS using virtual name in a cluster. Then install CI on a local drive on first cluster node. Install DI on a local drive on second cluster node. When logging in thru SAP GUI, you can either create a new entry to message server using Logon Groups or direct connection to CI/DI by using physical server name (just make sure to use correct system number).

Regarding your Dialog instance not coming up, try to stop and then start the instance from MMC. If still issues, look for errors in dev_server0, std_server0.out, std_bootstrap.out files located in "Work" folder on your DI.

I hope this helps!!

Fahad

[Former Member · Accepted solution]

Fahad,

Thanks for ur response.

I understood for the logongroups mentioned by u.... but what about the java part... like the 50000 page should open based on the local CI/DI instance number or by virtual ASCS/SCS instance number.

My dialog installation problem got resolved, actually i was installing the DI from the wrong menu in SAPINST. Choosed the correct menu or dropdown from sapinst and the problem got solved.

Also, a new problem has arised..... if start/stop the ASCS or SCS service from the windows services.msc rath than offline/online from cluster admin. The service resource fails completely and after that it never comes up.

Any idea abt the above problems....

Waiting for a reply asap.

[Former Member · Accepted solution]

Again, you have two options when connecting to Java.

Either you can use http://<physical hostname>:<Java port>/index.html to connect to individual servers (java port is usually 5<instance no>00)

OR

you can again use message server to load balance java servers for you by using

http://<SCS virtual name>:<message server HTTP port>/index.html (message server HTTP port is usually in 8000 series)

Regarding your second issue, you should always stop services/resources from a cluster. If you stop the service directly, then it will show as the resource failed in Cluster. But again, when you start it up from cluster, it should automatically start the service again.

Can you check when u start the service from Cluster, does it now start from services? If not, can u check for any application errors in Event viewer??

Fahad

[Former Member · Accepted solution]

Hi,

Its strange to see, I am able to connect to SAP instance through all the virtual hostnames.

like sapgui sapdbgrp 01, sapgui clusgrp 01..... connects successfully, whereas in ideal situation only sapgui sapcigrp 01 should connect successfully.

Waiting for a reply asap.

[Former Member · Accepted solution]

It's nothing strange with that.

On each node you have an application server.

The first node it's named the CI and on the 2nd it's named DI.

So it doesn't matter witch node you access, it will answer if it's started.

When you move the cluster group sapcigrp it only moves the enqueue server (SAP Locking mechanism) and almighty Message Server (who know everything about everybody...).

the message server will only answer to sapcigrp, but if all cluster groups are located on the same physical node it will answer also when you try sapdbgrp.

So:

In the cluster you have the SCS and the ASCS which you have to stop and start via the Cluster Administrator.

On each node, you have Application Server (CI/DI) which you have to stop and start via SAPMMC.

Picture this:

You shut down both cluster nodes and then start them, one after the other.

After this the users will not be able to logon to the system.

Why ?

beacuse you didn't start CI/DI via SAPMMC.

If you try to set the to Autostart=1, It might not work anyway since CI/DI needs the ASCS/SCS up and running in order to start.

A script with 30 seconds delay might do the trick here.

So, compare to NW04 (where the whole SAP system moved to the other node) things have become more complex.

This was done to minimize the Single-point-of-failure,

but have proven to be hard to understand for people who are used to a

"real" clustered installation where the whole application is in the cluster.

Puh, hope you'll understand what I mean.

[Former Member · Accepted solution]

Hi Tomas,

You said :

When you move the cluster group sapcigrp it only moves the enqueue server (SAP Locking mechanism) and almighty Message Server (who know everything about everybody...).
the message server will only answer to sapcigrp, but if all cluster groups are located on the same physical node it will answer also when you try sapdbgrp.

In my case, all the virtual groups are not in one node but the sapcigrp and clusgrp are in one node and sapdbgrp and msdtcgrp are on second node..... but sapgui still connects to all virtual groups at instance number 01.

Waiting for a reply asap.

[Former Member · Accepted solution]

In my case, all the virtual groups are not in one node but the sapcigrp
and clusgrp are in one node and sapdbgrp and msdtcgrp are on second node..... 
but sapgui still connects to all virtual groups at instance number 01.

When you via the SAPGUI specify hostname sapcigrp and instance 01, you are being directed to the node which holds the netname sapcigrp.

If sapcigrp is located on node A, the CI (instance 01) will answer.

If sapcigrp is located on node B, the DI (also instance 01) will answer.

Try with your SAPGUI to add two entries.

A - appl server=hostname of physical node A and instance 01

B - appl server=hostname of physical node B and instance 01

You will be able to logon on both of them.

So, on both MSCS nodes you will get an answer from instance 01.

It doesn't matter what the network name is.

As long as the instance is up and running, it will answer.

To get the grip, Assume that :

node A holds the sapcigrp and clusgrp and

Node B holds sapdbgrp and mstscgrp.

Then logon to the Node B and stop instance 01 via SAPMMC.

via SAPGUI you will now fail to logon with hostname sapdbgrp and mstscgrp.

If you move SAP Cluster Group to node B, only clusgrp is left on node A.

Thus, via SAPGUI only network name clusgrp will answer

(the actual hostname of Node A will answer as well).

Start instance 01 via SAPMMC on Node B and everybody are back to business again.

[Former Member · Accepted solution]

All,

When I go to the SLD page..... Administration-> Content Maintenance-> Subset:Landscape Description Class:HTTP Service Port(XX)... it gives me a list of SAP_HTTPServicePort.

When I click on any of the links for eg:- Port for AdminTool of af.<SID>.<Virtual Name> to check the details inside....I find the SecureURL: & URL: points to physical hosts (for eg: http://PhysicalHostname:5XXXXX/mdt).

How can i change to URL:http://<Virtual Hostname>:5XXXX/mdt permanently.

I tried changing at the same time and restarted SAP instances on both Nodes... but it seems that the SAP instance which starts last like either Node A (PAS) or Node B (AAS)... the URL: changes to that physical hostname:5XXXX/....

Please help.

Below is the enviorment:

Homogenous system copy (SAP PI 7.0) from win 2003 32 bit x64 (standalone with MSSQL 2000) to win 2003 64 bit x64 (MSCS with MSSQL 2005)

Source system export completed

MSCS Software installtion done

MSSQL 2005 software successfully clustered with failover

ASCS installtion successfully done

SCS installtion successfully done

Configure first MSCS node installtion successfully done

Install Database Instance successfully done

Configure additional MSCS Node successfully done

Installed Enque replication server for ASCS on both nodes

Installed Enque replication server for SCS on both nodes

Installed Central instance on first MSCS node

Installed DI-Dialog Instance on add. MSCS node

Waiting for reply.

Thnks in advance.