JSI Tip 10342. A long delay occurs before error recovery begins when a very large I/O packet is lost in a Windows Server 2003 SAN environment?

If a very large I/O packet is lost, or corrupted, in a Windows Server 2003 SAN (Storage Area Network) environment, it takes a very long time for the I/O time-out to occur, and for error recovery to start.

This behavior occurs because the Classpnp time-out algorithm erroneously ignores the 60 second maximum time-out in the registry, and increases the SRB (SCSI Request Block) time-out by 60 seconds for every 64 KB of data transfer requested.