Description of the Rolling Kernel Switch Concept
The rolling kernel switch (RKS) allows a new kernel to be imported into an ABAP system or specific parameter changes to be activated while the system runs.
To keep any adverse effects on business operations to a minimum when importing a new kernel version, instances are restarted one after another. This means that at least one application server instance is always available for business operations during the RKS procedure. The order in which the individual instances are restarted enables you to import a new kernel version without any system downtime.
Procedure
The system defines the order in which the application server instances are restarted. The SAP Start Service of the application server instance specified as the last instance to be restarted monitors the entire RKS procedure and triggers the restart of the individual instances (enqueue replication server, ASCS instance, application server instances). In turn the restart of the ASCS instance is performed and monitored by the SAP start service of the ASCS instance. So there are two monitoring SAP Start Services for RKS. The order in which the instances are restarted is described below. The figure shows the steps involved.
- The enqueue replication server is the first instance restarted.
- The ASCS instance is the second instance restarted.
- Then the application server instances are restarted in the order specified beforehand. In the figure below, instance A is the first application server instance restarted.
- Instance B is the second application server restarted, and so on.
- The instance defined as the last one is restarted together with its start service as the last of all C instances with the new kernel version. The RKS procedure is completed with this final step.

The ASCS instance contains both the message server and the enqueue server. All other application server instances are connected with the ASCS instance. The following section explains what happens in a system when an ASCS instance is restarted.
- The message server of the ASCS instance with the old kernel version (here version 1) sends a suspend request to all connected instances (suspends processing). Communication between the ASCS instance and the application server instances is briefly suspended. No requests get lost during the suspension period. Enqueue locks remain in place.
- As soon as all application server instances have interrupted the connection to the ASCS instance, the ASCS instance is started with the new ABAP kernel version (here version 2).
- During the restart process the application server instances regularly try to reconnect to the ASCS instance. As soon as the ASCS instance is restarted, the application server instances connect again to the ASCS instance. As soon as all application server instances are reconnected to the ASCS instance, the message server of the ASCS instance sends a resume request to all instances so that normal operation can resume.


