You may not immediately receive notification when a SATA hard disk fails in Windows 2000 Server or Windows Server 2003 (896065)

SYMPTOMS

Assume that a computer uses a Serial Advanced Technology Attachment (SATA) hard disk that belongs to an emulated Parallel Advanced Technology Attachment (PATA) hard disk master/subordinate drive pair. (A master/subordinate drive pair is also known as a master/slave drive pair.) In this configuration, you may not immediately receive notification when the SATA hard disk fails. Also, the computer may become unresponsive. If the drives are part of a software redundant array of independent disks (RAID) 1 pair, you may not receive notification of the failed mirror for several hours.

CAUSE

This problem occurs because PATA does not support surprise removal of devices. Therefore, the hardware cannot notify the operating system that a drive has been removed.

The operating system retries outstanding operations for the device until all the operations have timed out. The system may become unresponsive to requests during this time. This process may take several hours, depending on the number of outstanding operations.

RESOLUTION

To resolve this problem, configure the drives so that each drive is attached to a separate channel. Also, configure each drive to operate in master mode. When you do this, the operating system can test for the existence of drives on each channel. Then, the operating system can detect if a drive does not respond to low-level requests. If a drive does not respond to low-level requests, the operating system can remove that device.

MORE INFORMATION

PATA architecture represents drives within a drive channel where one drive acts as a master drive and a second drive operates in subordinate mode. The operating system communicates with the drives by channel. Therefore, if a drive cable is removed, or a drive loses the ability to communicate with the controller, the controller has no way to notify the operating system. Most of the time, the registers for the SATA drive controller indicate an ordinary operating state. The operating system continues to retry outstanding operations until they time out or until the controller indicates an error condition.