Web Dispatcher java load balancing

[Former Member]

Dear Experts,

I hope this is the right forum for posting my question.

We use a clustered MSCS ABAP+Java NW 7.0 system. Each cluster node has 1 dispatcher and 1 java server. Also, the SDM which is unique is either on one node or another according to where the SAP cluster group belongs. We also installed a SAP Web Dispatcher for load balancing. My question is this:

When the server in the 1st node fails (grey), but the dispatcher and SDM are still working (green) and in the 2nd node everything works (both server0 and dispatcher are green), shouldn't the Web Dispatcher pass all requests to the 2nd node?

I mean, shouldn't Java still be working fine? Isn't this the purpose of having a clustered Java instance?

In our case, if the server in the 1st node fails, Java fails. You can only logon to Java through the 2nd node URL.

The 1st node has a url http://server1.domain.com, the 2nd http://server2.domain.com. The url used for the dispatcher is the virtual ip's url http://server0.domain.com.

Thank you,

GS.

[martin_juen2 · Accepted solution]

hi.

you are right - normally you do not have to login to the second node if number one fails 🐵

can you tell us how your web dispatcher is installed and whixh ports are used? own installation, own directory, started from the kernel directory...

What i mean: your statement sounds that your web dispatcher uses the same http-port as server one. in this case the web dispatcher is no more available if server one fails and you have to make a new login on server 2...

I suggest you check the ports in your web dispatchers profile and change them if they are the same as used by server one.

we use web dispatcher which are installed as own system and also installed in an own directory - both versions work fine 🐵

regards, martin

[Former Member]

The problem was that the Java message server port was blocked. The ABAP message server port was 8100 (00 instance No), whereas the SCS was 8110 (10 instance No). Port 8110 was blocked by the firewall. When opened, clustering is working correctly meaning that when server0 from Node1 is down, users log on successfully to server0 of Node2 through the Web Dispatcher (or the virtual ip of the cluster if WD is offline). Thank you all!

[Former Member]

You are right stating that in a Java cluster, a Web Dispatcher is not needed.

We originally installed the Web Dispatcher only for redirect purpose.

However, forgeting the Web Dispatcher, let me state again my problem.

In the 1st node, java server0 fails (grey). The url used for accessing java is pointing to the virtual ip of the cluster and NOT in one of the 2 nodes. So it should work even if the java server in one of the nodes fails, right?

Well it doesn't. I cannot understand what is wrong. So any help appreciated.

[Former Member]

When you say Java in first node fails and is in Gray color you mean that the system did a fail over to the node 2 server?

If services from one node do not move to the node 2, it won't work because central services are still in node 1 and also virtual IP too. Remember that Virtual IP will move together with the SAP Group virtual name to the node 2 if a fail over happens.

If your services are moving then be sure the SAP Group has both nodes configured and not just one.

For the Web dispatcher you could enter http://yourservername:webdisp port/sap/admin and look for the servers the WD is connected to.

I have a similar configuration for some servers with ABAP+Java and the Web dispatcher works even if one of the nodes is down. By default it takes 120 seconds to switch (a parameter) and after this, the WD works flawlessly

[martin_juen2]

hi.

is your load balancing really working?

for me you have installed a scs instance and a central instance on server 1. on server 2 you have installed an additional dialog instance. when logging in you are working on the central instance (if no explicit logon on di happened).

how is your load balancing working / how do you tell the system that there are two servers to distribute the load and not only the central instance?

we have some portal clusters and in front of all systems a wwb dispatcher is installed to distribute the load...

regards, martin

[Former Member]

Hi,

Pls check the link..and check any thing that you have missed...

https://www.sdn.sap.com/irj/servlet/prt/portal/prtroot/com.sap.km.cm.docs/library/events/webinars/we...

Hope this helps you

Rdgs

Suresh

[Former Member]

The users Shold login to the SCS cluster name with the SCS http port.

When one of the node are down,

The users need to login again from the Cluster name and it will go to the other active node.

When one of the nodes fails it shold retrive only the working node. by refresh webdisp page. (login again)

but I think you dont need webdisp on JAVA MSCS

Thanks

Roy

Edited by: Royi Denis on May 17, 2010 10:47 PM