Install Enqueue Replication Server for MSCS

[ThomasKjelsrud]

Hi all!

I have just done a system copy within an MSCS environment, and also installed the Enqueue Replication Server (ERS). The problem is, however, that the installation of the ERS seemed to have gone wrong somehow. It is installed on a local disk on one MSCS node, but I only had the option to install in on a shared cluster disk on the second node. This caused the installation of the ERS to be incorrect. Can I, without risk, install the ERS again on the second node to see of I get the option to install it correct this time?

Regards,

Thomas

[ThomasKjelsrud]

Solved. I uninstalled the ERS instances, and installed them again. This time they got installed on the correct disks.

[Former Member]

HI

im new to this mscs can any one tell me what system do i need to pick up when installing ENQUEUE REPLICATION SERVER is it Netweaver system or ERP system.

regards

Raj

[MPGraziano]

Hi can you please advise the best method to uninstall a partially installed HA system

we have installed only centrall instance and first mscs node

[Former Member]

How about moving your shared disks to the other side when you perform the ERS install on node b?

[ThomasKjelsrud]

Hi Joshua!

Currently the shared disk is connected to node a, so in your opinion there should be no problem installing the ERS on node b? As I have already installed the dialog instance on node b, can I expect to get some issues? Will the instance number for the ERS on node b have to be the same as for the ERS on node a?

Regards,

Thomas

[Former Member]

Thomas,

I have not tested all possible scenarios but what you are proposing is supposed to work out of the box. There should be no problem install mulitple DIs or ERS on the same system or any combination as long as they use different system numbers. We run our ERS on 22 accross all of our HA SIDs. I have not tried running in different system numbers for the same instance. I am not sure if that is possible or would be successful.

Here is our setup though:

Node A

DB Normally running

DI running on 03

ERS running on 22

Node B

SCS normally running on 02

CI running on 03

ERS running on 22

Good luck

jwise

[ThomasKjelsrud]

Joshua,

thanks, I'll give it at try. I noticed that the ASCS and server cluster resources have to be restarted during ERS install.. so I guess it should be done during a production downtime window.

I have another query for you also. When I failover (Move group) the SAP Resource group from node a to node b, my disp+work.exe on the DI on node b refuses to start. I have configured the instance profiles on node a and be identically (except for instance number). Do you know if there are any extra things I have to do in order to get the DI to start when the failover happends, or should the DI on node b be running always?

Our setup is this:

Node A:

ASCS running 01

CI running 00

ERS 11

Node B

DI (not running) 02

DB running

The ASCS and DB are clustered. Do you have any recommendations to improve our setup, as it seems to be some misconfiguration here..

When I look at SAPMMC on node b, the disp+work is gray (stopped) but the elapsed time is still running. igswd.exe is green and running.

Thanks for you input!

[Former Member]

Hi again Thomas,

Yes, ERS requires a restart of SCS if done using SAPINST.

For simplicity's sake, I would recommend that your CI and DI both use the 00 number. Under the 04s architecture, the CI and the DI are now local concepts. They are designed to always run under normal operations. Your SCS should be in cluster, your CI and DI should be installed locally. SAP had made this change difficult and not obvious in some cases so no one would blame you for the confusion.

The SCS is the HA component that allows the message server and to run nearly non stop and ERS is the HA component that allows the enqueue service to run nonstop by providing triangular lock table replication amongst the SCS and 2 or more ERS services. As long as the lock table is preserved in this fashion, the users session remains persistant. Not having ERS, as you may have observed, causes the jcontrol to issue a restart to all server processes in the event of an SCS failure or failover. Best practice is to have ERS running on Node A and Node B. In your scenario, if you lost Node A with SCS running, you would loose your instance.

The disp+work being gray could be any number of things. The first thing to check is that the MMC is not reporting it wrong. Testing a SAPGUI connection directly to the DI would tell you if the DI is actually running, of course so would the task manager. If it is running then I would recommend a sapstartsrv.exe reinstall of the service.

If it is not running, it could be a number of things. Start with the disp+work trace, dev_w0, to see what is happening. You can post if you want some advice but the complexity could be such that I would not be able to help without troubleshooting.

[Former Member]

Hi Joshua,

I have installed our NW Portal 7.0 (AS Java, EP Core and EP useage types) using the same distribution as you have stated below.

There is no mention of Web Dispatcher in this scenario and i have not installed it myself either....is it required?

Am i right in saying that if the SCS instance is running on Node B, it will use the Application Server running on Node B, but if the SCS fails over to Node A, it will use the Application Server running on Node A.

What happens if the Application Server on Node B (installed locally) fails while the SCS is also running on Node B? Will the SCS automatically detect this and start using the App Server on Node A...or do i have to manually fail the SCS over to Node A first?

Also, after i have installed my Diallog Instance (additional Application Server), there are missing directories in the '....../j2ee' directory....e.g. Visual Administrator and JSPM files....is this usual?

Regards,

Sean.