Sap executables corrupted by failover in MSCS

[Former Member]

We installed the an NTW04s ABAP+JAVA in MSCA.

So on the first node is instalelled (locally) the CI, while on the second node is installed one Dialog Instance.

On both the nodes we installed the ASCS + SCS services, and for all of them, all both the nodes, we installed

the relative Enqueue Replicator Servers.

After the installation we noticed problems doing the move group for the SAP group with the Cluster Manager.

The ASCS and the SCS resources go in "Failed" status on the other node, and did not come up.

Also to move back the SAP group do not solve the problem, as the ASCS and the SCS resources continue to staty in

failed status.

What we noticed is that as the move group action is started there is a message coming from the operating system that sound something like:

" an error occured on the X: drive..trying to write (or save) the

sapstart.env (or sapstartsrv.exe) file...data has been lost..."

After that the only solution to make the ASCS and the SCS services able to start again is to replace the sapstartsrv.exe on the global sapmnt with a backup copy.

The sizes of the these files is not changed due the move group action, by the way replacing them with a backup

copy allows the services to start again.

So it seems these executable are corrupted during the failover.

We are using the 133 pl kernel.

We need to clarify if this behaviour can be caused from disk errors on the MSCS nodes or instead from a missing Sap patch level.

Someone experimented this strange error ?

[Benny · Accepted solution]

Marco,

installing SAP systems under MSCS is clearly described in the install guides. You must follow these instructions especially regarding the server setup. Please go through this again and check whether there are any discontinuities.

Regards,

Benny

[Former Member · Accepted solution]

Tnks for the feedback.

have to fix my first description of the scenario.

The ASCS e SCS are installed on shared disks, not locally on both nodes.

On both nodes we run two times the installation of the ENqueue replication

service, the first time indicating the ASCS, the second time the SCS instance number.

In short we did the installation according the Sap manual.

On both nodes the Enqueue Replication services have the same system numbers.

The Primary and the Secondary application servers (Abap+Java) are installed locally on different nodes and uses different sistem numbers.

Any idea why the sapstartsrv.exe should be corrupted by a failover of the SAP cluster group ?

[Former Member · Accepted solution]

How did you resolve this issue?

Regards

Chandu

[Benny · Accepted solution]

Roberto,

please issue an OSS message. This sounds weird and maybe an error.

Regards,

Benny

[Former Member · Accepted solution]

We had a similar scenario and were able to correct it by using the following SAP notes 1043592 & 867521.

Hope it helps.

~Jamie

[Former Member]

applied patches attached to note 1043592

[Former Member]

Hi Marco,

There is nothing to do with patch level, it seem's there is some issue with services only. If you don't move the services in the consistant mode( closing the instance and stopping the services) this kind of problems will arise.

Regards,

Hari.

[Former Member]

please this is an MSCS scenario.

IF one node get down the cluster resorces should be passed on the other node and then restarted automatically.

Otherwise what's the reason to exist for an MSCS scenario like this used by Sap ?

regards

[Former Member]

Yes , it will work. Please go through the installation procedure again...