SCSI Reservation

SCSI-2

• Exclusive reservation; only one node can own it.
• Does not survive node reboots; not persistent.
• One path with one host.

SCSI-3

• Persistent (group) reservation.
• Persistent across SCSI bus resets or node reboots.
• Supports multiple nodes accessing a device.

SAM provides two methods of error recovery from an ACA condition, which is determined by the Normal ACA (NACA) bit in the control byte of the SCSI command.

Ø  If an ACA condition occurs for a SCSI command that has an NACA value of 0, then an ACA condition occurs that requires SCSI-2 contingent allegiance error recovery to be used.

Ø   If an ACA condition occurs for a SCSI command that has an NACA value of 1, then the new SCSI-3 error recovery (as specified in SAM) is used.

The primary advantage of the new SCSI-3 error recovery is that a SCSI device's queue is frozen until it receives the Clear ACA task management function. After a SCSI target receives the Clear ACA task management function, it will either flush its queue of pending commands or resume execution of them, depending on the value of the Qerr bit in the control mode page (mode page 0xA).

FCP (SCSI on FCS) requires all SCSI devices to automatically return Request Sense data whenever an auto-contingent allegiance (ACA) condition occurs. This is referred to as autosense.