Fujitsu MAS3735, MAP3367, MAP3147, MAP3735, MAS3367 Init Error Recovery Methods Recommended

5.2 INIT Error Recovery Methods (Recommended)

(2)RESERVATION CONFLICT status

This status shows, under a multi-initiator environment, that the IDD is currently reserved by another INIT, and cannot be used until that reserved status is released.

The period over which the reserved status continues should be managed by the system, but normally, an INIT that receives this status reissues the original command after waiting an appropriate length of time. The wait time until a command is reissued is generally an inherent value of the system and it is necessary that consideration be given to the operating states of the other INITs in the system when deciding it.

Of the INIT which reserved the IDD falls into a state which it cannot recover from for any reason before the reserved state is released, it will be impossible for other INITs to access the IDD because that reserved state will continue. In such a case, it is necessary for the INIT to take one of the following measures to forcibly release the reserved state and recover the access right.

•TARGET RESET message

•RESET condition

(3)Command completion wait time-out

This error is a state where the completed status for a command is not reported within the time expected by the INIT that issued the command.

One cause of this error occurring, besides a malfunction of the IDD, is that the command was cleared by some method which the INIT that issued the command cannot clearly detect, such as a TARGET RESET message, or the command was cleared when an unrecoverable error was detected in reconnection processing executed by the IDD, etc. When this error occurs, the INIT should gather sense data by issuing the REQUEST SENSE command, then judge which recovery method to use based on the contents of the sense data (see Section 5.2.2).

Not only does command execution time differ widely depending on the type and specified contents of the command, but since it has an influence on the following factors, it is necessary to give adequate consideration to these points when setting the monitoring time value at the INIT.

•Since the command stack feature exists (see Section 1.4), even if the IDD accepts a command normally, if a command issued by another INIT is being executed, or if there are other commands in the stack, execution of the command is caused to wait until all the other commands have been processed. The execution wait time is dependent upon the type of command issued by the other INITs and the contents of its specifications.

•If the IDD has executed error recovery processing (retried) for an error that occurred on the SCSI bus or in the disk drive, command completion time will be longer than normal.

(4)BUS FREE error

This error occurs in the case that the SCSI bus enters the BUS FREE phase at a time when the INIT is not expecting it, during execution of a command on the SCSI bus. Even in cases where a RESET condition is generated by another SCSI device on the bus, the same state occurs, but the occurrence of an unexpected RESET condition should be processed as an abnormal state of the system.

C141-E167 5 - 13