Abstract: An embodiment of the invention is a technique to manage failover and failback. A failover of a first path is detected. The first path corresponds to a first device in a plurality of physical devices having M device types. A connection status of the first device is determined if the failover is detected. The connection status is one of a connected status and a disconnected status. The disconnected status corresponds to the failover. The first path is adjusted according to the connection status.

Abstract: An embodiment of the invention is a technique for maintaining application availability during a change in a resource dynamic link library (DLL). A shim resource DLL is linked to a resource DLL managed by a resource manager in a clustered system. The managed resource DLL exports a function to control a resource in the clustered system. During a normal mode, the shim resource DLL passes to the managed resource DLL a call to an exported function unchanged. During a change mode, the shim resource DLL is unlinked from the managed resource DLL to allow the managed resource DLL to be changed, and the shim resource DLL handles a call to an exported function without passing the call to the managed resource DLL.

Abstract: An embodiment of the invention is a technique to manage multipaths for input/output (I/O) devices. An I/O request packet (IRP) from a higher level driver is received. A plurality of paths to a plurality of device objects is managed in response to the IRP using a plurality of lower level drivers. The device objects correspond to physical devices having M device types. The lower level drivers control the physical devices.