DS20E IO BUS RESET PROBLEM

[Mrityunjoy Kundu]

Hi,
In our territory there are four alpha systems(one ES45,two Es40,one DS20E)which are in cluster,where openvms7.3-2 is running.When we reboot DS20E server,during the time of booting, scsi bus reset is detected and it reflects all servers console and errorlog.But, when we reboot other servers ,during the time of booting this problem doesnot occured.why??

[Duncan Morris]

Mrityunjoy,

you really need to give us a bit more information!

What is your SCSI configuration? What adapters are in use by each system, and what else is on each SCSI bus?


Duncan

[Hoff]

SCSI bus reset operations are a normal part of multi-host SCSI, and a central reason why HP does not support tape drives on shared buses. (The reset causes the drive to rewind, regardless of what might have been occurring.)

SCSI bus resets can trigger an I/O retry, but should not and do not trigger application-visible problems.

As for why some systems do and some systems do not reset the bus, you can influence this console behavior with the SCSI_RESET console environment variable. (SCSI_POLL is also related to the initialization of the buses.)

I generally prefer to reset the buses, as the results are rather more predictable. Which means I regularly see SCSI resets. And I then regularly ignore the logged resets whenever I know another host rebooted. I do get interested when resets arise when no host rebooted. This usually means hardware issues, bus issues, or somebody was (erroneously) hot-plugging SCSI drives.

Do also ensure you have a SCSI controller that is supported for multihost operations. Per HP, some of the controllers in widespread use are simply not supported for multihost operations. These configurations can potentially show up with bus resets.

Lists of these controllers HP supports for multihost operations are available from HP. Use the google "site:h71000.www7.hp.com" keyword and google to dig around for this and for OpenVMS info available from HP.

Stephen Hoffman
HoffmanLabs LLC