SAP High Availability after Microsoft server node restart

[AndreC1]

Hallo everyone,

Is there a way to automatically start SAP DI instances in an ACTIVE/ACTIVE Windows ASCS cluster when one of the nodes becomes available after it was offline?

Here is what happens in our environment (Windows 2008 R3, MS SQL 2008 R2, ECC6 EHP7). After one of the servers goes offline (say node 2) and it returns to service, it does not start the SAP DI (disp+work.exe). We have to manually do that from the SAP MMC. This means that after a server (node 2) returns from an offline event such as a restart, SAP is no longer in a HA environment. Should the first server (node 1) go offline, there is no DI started on the second server (node 2) to support end users. Is this the expected default behavior?

Is there a way that when a node restarts, that it can check to see if the ACSC is active, and if so start the DI? If not, then stay off line. Or is that at least a way to notify the administrators that the system is no longer HA?

[Former Member]

I hope you should be able to add the relevant services/resources to the cluster setup so that the DI (dialog instance..) can automatically start without manual intervention whenever the physical node comes up

but before this the ASCS/DB should be up and active in the cluster.

regards,

Chandru

[Sriram2009]

Hi Andre

1. In both Windows MSCS cluster nodes SAP instance kept on. when the system performing failover to node 2 it will start automatically.(it mean that any failure in the node 1 and during this process both nodes SAP instance should be on)

2. How you are making the node 2 goes offline? by manually or automatic?

3. Kindly refer the Windows MSCS FAQ link which will address all your query

Regards

Sriram

[AndreC1]

Hi Sriram,

Sorry for the delay. I initiated the failover by restarting the actual server remotely. I wanted to simulate a node failure. I have read the FAQ on High Availability, but it does not address this issue from what I can see.

What I see is that when both servers are running and we started the DI on both systems, the CI will transfer between the two nodes and SAP will continue to be available. But, if you recover from a restart, the SAP DI does not start automatically on the server that restarted. When you then move the CI, or the CI fails from the working node to the node that came back from a restart, the CI will move to the restarted node, but SAP will be unavailable to the clients, because the DI did not start automatically. Is this normal behavior?

[Sriram2009]

Hi Andre

What I see is that when both servers are running and we started the DI on both systems, the CI will transfer between the two nodes and SAP will continue to be available. But, if you recover from a restart, the SAP DI does not start automatically on the server that restarted. When you then move the CI, or the CI fails from the working node to the node that came back from a restart, the CI will move to the restarted node, but SAP will be unavailable to the clients, because the DI did not start automatically. Is this normal behavior?

Once the physical cluster node restarted then you have to start the SAP DI instance manually. I think it’s possible to do the auto start but it not sensible in case of any issue in the physical server,

BR

SS

[AndreC1]

Thank you Sriram. That is very concerning then. In reality it means that in such a case, SAP is only HA for the first fail over. Should the second physical node go off line before the administrator manually intervene, SAP will become unavailable. Even further, if the administrator does not realize that the first node DI is off line, then SAP could be in a non-HA state for days.

Any ideas on how to setup a notification system to alert the administrator that the DI is offline?

[Sriram2009]

Hi

You can use the both nodes active mode. one of the node act as CI for SAP application.another one can handle the Database. all the end user activity are in CI & admin's/ function guys are in DI.

I am not sure about the alert notification in windows failover. you can check this thread
Cluster Failover Notification - kb260527.

BR

SS